Privacy Policy
Last updated: 16 November 2026 · Burncaps is operated by Jordan Labs Ltd, a company registered in England and Wales (company no. 17358646).
The short version
We never see your videos. Your video is processed only on your device and is never uploaded. The sound goes only to our speech service: on a phone it does unless you switch it off, and on a computer if you choose it (for a language other than English, for a recording over ten minutes, when you ask for a closer transcript, or, on Pro, to place a podcast episode’s clips in your own file of it), or by itself if the computer is too slow to listen well (the page says so, and one press keeps the sound on the computer). Then only the sound, never the picture, goes through our server to Microsoft’s speech service in the EU, which by Microsoft’s terms for this service stores none of it, and we store none of it either. One more case, only on Pro and only if you add your podcast’s public feed in your account: the same service hears each new episode’s published sound, from the address in your feed, never your own file, apart from a few short pieces of its sound when your podcast host adds adverts to the published episode (New episodes of your podcast). We also see how the site itself is used, through PostHog, an analytics service on its EU cloud: the main steps a visit takes, and, outside the European Economic Area, for some pages, a recording of how the page changed, with your video, your sound, your words and everything you type left out. Nothing is stored on your device for it (How the site is used).
What we process
Your video: processed entirely on your device. Never transmitted to us or anyone else.
Your sound: it stays on your device, except that it goes to our speech service: on a phone unless you switch it off, on a computer if you choose it or if it is too slow to listen well (below).
Transcripts and captions: stored on your device only, except the transcripts of the episodes of a podcast feed you add on Pro, which our server keeps for 30 days (below).
Your logo and brand presets (Starter): a logo you choose is read on your device and kept in your browser’s own storage (IndexedDB), and your brand presets (their names, the caption style, colours, position, size, words at a time and which logo) and the logo’s corner and size are kept in its local storage. They are drawn onto the video on your device, like the captions, and never sent to us or anyone else: no request carries them, and the step counts say nothing about them. Clearing this site’s data in your browser removes them. One exception, only on the Agency plan: a client’s look (the caption style, colours, font, position, size, words at a time, opening title, frame, ending, background, clip length, house hashtags and names to spell) is kept with your team’s account so your teammates share it (see “Agency” below). The client’s logo file stays in the client’s folder; only which logo, its corner and its size are kept with the look.
Our speech service (on a phone unless you switch it off; on a computer by choice, or by itself when it is too slow to listen well): on a phone, the words are written by our speech service without a question first, because a phone’s own speech model is slow and often runs out of memory. While it listens, one line above your video says so, with a button, Keep the sound on this phone: tap it and nothing more is sent, the phone listens from the start, and your browser remembers the answer for every later video (change it back under Listening settings). A phone that keeps its sound is asked once before a recording over ten minutes is heard, showing how long each way takes, and once while it listens if the phone will take more than another minute; nothing is sent unless you choose our speech service there, and the answer your browser remembers stays as it is. If you keep the sound on the phone and a video would take more than a minute more, the page asks once whether to use the service for that video. Ticking Keep everything on this device sends no sound at all, on a phone or a computer. On a computer it asks when your video is in a language other than English (the same remembered answer), before a recording over ten minutes is heard (once, showing how long each way takes; your browser remembers that answer for later long recordings, and Listening settings or the line above your script changes it), when the computer is listening slower than the video plays but would finish within three minutes (asked once for that video, and not at all once you have kept the sound on the computer), and it offers a button to listen again with the service for one video. A computer too slow to listen well uses the service by itself, with no question first, as a phone does, and only while your allowance on the service holds the recording. That is read from what your device does, never from its name: the browser reports no WebGPU graphics and four or fewer processor cores or 4 GB or less of memory, and the computer has not already listened at least as fast as a video plays (the speech model is then not downloaded at all); or its pace in the last 7 days, kept in your browser with its date, says this one would take longer than it plays and more than three minutes; or, once it has heard a minute of sound and two pieces of it in a row were slow, its pace says the same (it then starts again with the service, so no word is lost or heard twice). One line above your video says so, with a button, Keep the sound on this computer: press it and nothing more is sent, the computer listens from the start, and your browser remembers the answer for every later video, recordings over ten minutes included (change it back under Listening settings). With the service, the page cuts the sound into pieces of up to 30 seconds at the speaker’s pauses and sends each piece, as a 16 kHz mono recording with nothing from the picture, to our server, which passes it to Microsoft’s MAI-Transcribe (Azure AI Speech fast transcription, North Europe region, Ireland) and returns the words with their timings and the language heard. On a phone, when the connection drops or the service is busy, the page waits a little and sends the pieces still missing again, cut smaller (up to 10 seconds each), up to three more times over about a minute, before the phone listens on itself; Keep the sound on this phone stops that too, at once. Microsoft states that for fast transcription it “does not retain or store the data provided by customers”. Our server keeps no sound and no words: it counts the seconds heard for a random ID for your browser (or your email when signed in) and for your connection, stored as salted one-way hashes for 30 days, so the allowance is fair (three hours a month on Free, 5 hours with Starter; after that the words are written on your device). It also adds each piece’s seconds to daily totals by the language heard, phone or computer (from what your browser says it is), plan (free, trial or paying) and speech service; those totals hold nothing that points to you, your browser or your connection. Before a recording over ten minutes is heard on a computer (unless you chose to keep long recordings on the computer) or on a phone, before one goes to the service from a Starter account, and before a computer too slow to listen well sends any sound, the page asks our server how much of the allowance is left, for the random ID for your browser or for your account when signed in, so it offers the service only when what will be heard fits and tells you first when it does not; asking counts nothing and sends no sound. Tick “Keep everything on this device” and no sound is ever sent.
B-roll (Pro and its free trial; added by itself the first time you open a clip with Auto edit on, or when you press Add B-roll; switch B-roll or Auto edit off and nothing is sent): the words of that clip as heard, with their times, the names you typed and the language go to our server and on to Claude (Anthropic), once for that clip, which names up to five moments where you mention something a viewer can picture and a few plain words to search a stock video library for each. The page also tells our server whether it is on a phone, so the shots come at a phone’s size. Our server sends only those search words to Pixabay, every one from a fixed list of everyday English words (never a name), with our own key, and reuses Pixabay’s answer to each search for 24 hours, as Pixabay asks; nothing about you is in it. Your browser then fetches the stock shots you see, and their small pictures, from Pixabay’s own server (cdn.pixabay.com), which sees your connection’s address and browser as any website you load a picture from does, and draws them into your clip on your device. Your video and its sound are never sent anywhere for this. Each request is counted on your account (a few a minute and a few dozen a day), stored under a one-way hash of the account with the day and the moment of the request, and deleted after two days; the same is kept, and deleted the same way, for each time our AI looks for clips for a Pro account; without Pro or its trial, or with Keep everything on this device ticked, nothing is sent at all.
Your projects: every file you open becomes a project in the list at burncaps.app/app. The file, the script made from it and the clips found in it, with your edits to each, and a small picture of one frame are kept in your browser’s own storage on your device, so that going to another page, closing the tab (or a phone closing it for you) loses nothing, and you can start another video and come back to this one. The file you have open is copied there whatever its length, whenever your browser says it has room for it; older copies make way to keep them all within about 10 GB of copied files on a computer and half that on a phone. A file there is no room for is not copied: only its name, size and length are, beside its script and clips, and the list asks you to choose the same file again (the tab you opened it in never asks). While our speech service is hearing a file, the parts of its script already back are kept the same way for two hours, so that if the page is closed partway, choosing the same file again carries on from there instead of sending those pieces again. They never leave the device. On an Android phone, in Chrome or a browser built on it such as Samsung Internet or Edge, the part of a video already written while it is exported is kept the same way (the finished picture, a few seconds at a time, with where it stopped), your browser’s local storage notes how far that part reaches and, during an Export all, which of its clips are already saved, so that if the phone closes the page partway, exporting the same video (or Export all) again with the same settings carries on from there instead of starting over. The written part is deleted when the export finishes, is cancelled or fails, the Export all note once every clip of it is written, and anything of either still there is deleted on your first visit to burncaps.app after two days. If an export stops because the browser could no longer read your file, the written part is kept, so choosing the same video again carries on from it. They never leave the device either. On a computer, a tablet, an iPhone or an iPad none of this is kept. Up to 24 projects are kept, none longer than 60 days; past that room the oldest copies are removed (their scripts stay); “Delete from this device” on a project removes it at once, and clearing this site’s data in your browser removes them all. A project from a connected client folder (Agency) is read from that folder and not copied. A whole video over ten minutes is written, as it is made, into a file in your browser's own storage on your device when that browser is Safari or Firefox, and when it is Chrome on a phone short of room for its usual way of keeping a file. It never leaves your device. Once its download has had time to finish it is deleted, about 15 minutes after a download that starts by itself and about an hour after you tap Download, or at your first visit or next video that long after that. A file left half written by a tab that closed is deleted at your next visit. A finished video you have not saved yet on an iPhone or iPad is kept for up to three days, so that a closed tab does not lose it, and the page offers it when you come back. A file that failed is deleted at once. The notes in this browser that keep track of these files are bc_disk_out and bc_disk_ready.
Agency (a team of five people): Agency is not on sale yet; this is how it will work when it opens. Each client’s recordings stay in that client’s own folder on your computer (synced by your Drive, Dropbox or OneDrive if you use one). Burncaps reads them from the folder in place: they are never copied into the browser and never sent to us. Burncaps writes only into a folder you choose in your browser’s own folder window, and only inside Burncaps clips: the clips you export, a posts sheet for each episode (each clip’s file name, title, caption, hashtags, length and start), the script of each episode (its words and their timings, with the clips found and their titles and posts), the client’s logo, and a small file holding only an id, so a teammate’s copy of the folder is known as the same client. Which folder is connected is remembered in your browser (IndexedDB) and never sent. Our server keeps, for your team, through requests to /api/team: the email of each person on it and whether they pay for the team; for each client its name, the connected folder’s name as your browser shows it and that id, its look, house hashtags, names to spell and fixed words; and for each episode its file name, size and length, its status (new, clips found or exported), who last saved it and when, and the clip choices as word positions and cut marks, with which clips were exported, by whom and when. It never keeps the video, the sound, the script, a clip’s title or a post. The sound is heard as everywhere else on Burncaps (see “Our speech service”). Finding clips in new episodes by themselves (on for each client unless your team turns it off): while Burncaps is open on one of your team’s computers, it looks in each connected folder your browser already allows it to read, when the tab opens, when you come back to it, when someone turns a client’s switch on and every 10 minutes, and never asks for a folder by itself. It takes only recordings added after the folder was connected or the switch was turned on (your browser notes the names and sizes of those already there, on that computer, and never sends them), at most 3 at a check and 30 a day for the team; Find clips takes the rest. A new episode is heard and its clips found on that computer exactly as when someone clicks Find clips, one at a time, and it never starts one while something is being edited or exported there. A recording over ten minutes is not heard by itself until someone on that computer has chosen where recordings that long are heard, on the computer or by our speech service; the Clients view asks. For this our server also keeps whether each client’s switch is on and since when; who last checked the folders and when; while an episode is being heard, whose Burncaps is finding its clips, a random id for that browser tab and when (so two teammates’ Burncaps never take the same episode by themselves, and anyone opening it by hand is asked first; it lapses after 30 minutes); how many episodes the team’s Burncaps took by itself each day; the episodes the emails announced or will announce, kept for 2 days; when each person was last emailed and how many emails the team had that day; and whether you stopped these emails. When clips are found this way, each person on the team who is signed in on a device and has not stopped them gets one email at most an hour, covering every client, and the team at most 20 a day, carrying only each client’s name, the episodes’ file names, how many clips were found, a button that opens Burncaps at those clients and a link that stops these emails for that person (turn them back on under Clients). An email waits while 60% of our day’s email allowance is used, so sign-in codes and account emails always go first; one held back goes at a later check by a teammate’s Burncaps, and is dropped if it has not gone within 2 days. Remove a client and its records go at once; when the plan ends they are deleted after 60 days, with the team. Each person on the team signs in with their own email and a code, on up to three devices.
“Tell me when ready”: if you switch it on, your browser shows its own notification when your script is ready or your export is written while you are in another tab. The notification is made on your device by this page; nothing is sent to us or anyone else to make it.
A file you drop on one of our other pages: the search, language and translate pages have the drop box too, and the safe zone checker has a “Caption this video in Burncaps” button, the video length checker a “Cut it into clips in Burncaps” button, the LUFS checker a “Caption this video in Burncaps” button, the speaking speed test a “Measure a recording in Burncaps” button, the video trimmer, the MOV to MP4 converter, the audio extractor and the video compressor an “Open this video in Burncaps” button, the video volume tool a “Caption the louder video in Burncaps” button (it hands over the louder file it made) and the video size checker (on the Instagram Reel, YouTube Shorts and TikTok size pages) a “Make it 9:16 with captions in Burncaps” button (4:5 or 1:1 if you pick that shape, or “Cut it into clips in Burncaps” for a video over ten minutes), which first saves the shape you picked as the editor’s Shape setting in your browser’s own storage, but the editor is on the front page, and a browser cannot pass a file from one page to the next. So a file you drop or choose there is kept in the same storage on your device, the tab goes to the front page, and the front page opens it from there. It never leaves the device. Only one is ever kept: the next one replaces it, and the front page removes it the next time it loads. If your browser will not keep it (a private window can refuse), nothing is kept and the front page asks you to drop it again.
AI model files: your browser downloads the open-source Whisper model from Hugging Face's content delivery network and, when a video is in one of twelve Indian languages (Hindi, Urdu, Bengali, Tamil, Telugu, Gujarati, Marathi, Kannada, Malayalam, Punjabi, Nepali or Odia), the open-source IndicConformer model made for that language, from the same network. The code that runs these models comes from the jsDelivr content delivery network, as do the face and person-outline models used for framing and backgrounds, and the demo video, the example clips and the sample video on the home page. These downloads are normal file requests and include no data about your media.
New episodes of your podcast (Pro, only if you add your feed): on Pro you can give us the address of your podcast’s public RSS feed in your account, after ticking that it is your show or one you may make clips from. Our server keeps the address and the show’s title, and reads the feed at most once every two hours, asking each time only whether it changed; every request names Burncaps and links to this page, and your podcast host may count it. When a new episode appears (and once for the latest episode when you add the feed, if it came out in the last 30 days), our server follows the episode’s sound address from your feed, reads the published sound once into its memory, measures its length from the sound itself, and sends it to Microsoft’s speech service in the EU (the same MAI-Transcribe as above, which by Microsoft’s terms stores none of it), which writes the words with their times; our server keeps no copy of the sound. The words with their times then go to Anthropic’s Claude API to find and rank the clips, exactly as the clip finder above, counted on your account. Your own file is never sent: when you drop your file of the episode on burncaps.app, on a computer or a phone, the page asks our server, with your sign-in, which of your ready episodes have that length, or are 5 seconds to 20 minutes longer and were ready in the last 14 days (it sends the length and nothing else about the file), decides on your device from the length and the file’s name which one it is, asking you once when it is not sure, and then fetches that episode’s transcript and clips, so they appear with no listening. When your podcast host adds adverts to the published episode, your file is shorter than what our server heard, and every word after an advert is later in that transcript than in your file. Then, to place the clips in your file, the page sends a few 20-second pieces of your file’s sound (never the picture) to our speech service, the same way as any listening above and counted in your speech hours (at most 24 pieces, 8 minutes, for one file), matches their words with the episode’s transcript on your device, moves every word and clip to your file’s time and leaves out the adverts’ words; if they do not match, it says so and listens to your file as usual. A phone does this unless you keep the sound on your phone; a computer asks you first, unless you chose our speech service for recordings over ten minutes; with “Keep everything on this device” ticked it never happens. For each episode our server keeps, for 30 days: its title, number, date, length and sound address, the transcript with each word’s time, the clips (title, opening card, score, grades, the reason, which words each one keeps), whether you opened them, and the episode hours heard, which count against the 20 speech hours of Pro and an 8-hour limit for episodes in any 30 days. After 30 days the episode and its transcript are deleted. When an episode’s clips are ready we send you one email, to your Pro email through Resend; every one has a link that stops them, and your account has the same switch. In your account you can delete one episode, remove a feed (its episodes go with it) or delete everything at once; when Pro ends, the feeds are no longer read and the episodes are deleted when their 30 days end.
Payments: handled by Stripe. We never see your card details. Stripe shares your email and payment status with us so we can switch your plan on for you. If a payment fails, Stripe tells our server that it failed, with the amount and currency, the reason, the country of the card and the email used, and we keep that in our log so we can help you pay. We then email the address you gave at the checkout about that payment and nothing else: for a first payment, once, up to about 90 minutes later and only if you have not paid by then; for a renewal, with the link to pay it, at most three times before your plan would switch off, and not once you have cancelled it or replaced it with a new subscription. When you are signed in here (Pro, Starter, a free email or the trial), the link to Stripe's checkout carries your email so the checkout is filled in with it; it goes to Stripe only. See Stripe's privacy policy for how they handle payment data.
Subscriptions (Pro and Starter): handled by Stripe, which holds your card, your billing details, the subscription and its invoices, and emails you about payments. Stripe tells our server about your subscription as it changes (started, renewed, a payment that failed, cancelled, ended), and our server keeps, with your sign-in record below, the subscription’s and Stripe customer’s reference numbers, its status, when the paid month ends and whether it was cancelled, so it can switch Pro or Starter monthly on and off. You manage or cancel the subscription on Stripe’s own billing page, or by emailing us. On a yearly plan our server also sends one email about a week before each renewal (its date, its amount and your billing page), read from the subscription record above; it keeps only that the email for that renewal was sent.
Your sign-in for Starter or Pro: both are tied to the email you paid with. Our server keeps that email, a record of the payment, and up to three signed-in devices (ten with Pro) (for each: a random identifier, a short label such as "Chrome on Mac", when it signed in and when it was last used). Each device holds a random sign-in token in its browser storage; our server keeps only a one-way hash of it. When you sign in on a new device we email you a six-digit code, sent through Resend, our email provider; the code is stored only as a hash and expires after 10 minutes. If that email has no account yet, we keep only salted one-way hashes of it: beside the code’s hash and the times codes were sent, beside a count of wrong codes typed for it, and beside the time we last found no purchase for it at Stripe; the account is made only when the right code is typed, and all of these are deleted within three days. An email that has not signed in is sent at most three codes a day. If a sign-in is asked for with an email whose purchase was refunded or disputed, no code is sent. Where an email that never bought would be sent one (to start the free trial, at an export, or for a free account), that address gets one short email instead, saying there is no plan to sign in to; any other sign-in with it sends nothing. We also count sign-in attempts per internet address each day (for an IPv6 address, per network and per wider network), stored as a salted hash, so nobody can guess codes. If you answer the one question on the page Stripe returns you to (where you first heard of Burncaps and, if you like, what you caption, each a tap on a short list, with up to 80 characters of your own words for “Somewhere else”), the answer is kept with this record. We use your email for nothing else: no newsletters, no marketing, unless you ask for our emails yourself (below). Ask us to delete it at any time.
Continue with Google (since 11 November 2026): where you sign in or start the free Pro trial, a “Continue with Google” button can replace typing your email and a code. No Google script runs on our pages: the button is ours, and only when you press it does Google’s own sign-in open (accounts.google.com), in a window on a computer or a new tab on a phone, asking for your email address only. Nothing goes to Google before you press it. If you choose an account, Google sends you back to our page burncaps.app/google-connect.html with a signed note saying which email address it has checked is yours, in the part of the address that is never sent to any server; that page passes the note, sealed, to your Burncaps tab and closes, and the tab passes it to our server, which checks Google’s signature itself (fetching only Google’s public list of signing keys) and then signs you in exactly as a correct code would. We keep only that email address, on the same record and under the same rules as a sign-in with a code (above); your name, your picture and your Google account ID come with the note and are thrown away at once, never kept. We do not get your Google password, your contacts or anything else in your Google account, and we never post or write anything to Google. The note itself is not kept.
The free Pro trial (7 days, no card): it is offered on the home page, with the plans, on the Pro page and in the app before you reach the limit of Free (not on the card shown at the limit) and started with your email only. Before it starts, our server checks the part of your email address after the @ against a list of throwaway email services kept in our own code (no request goes anywhere else); since 2 November 2026 a throwaway (disposable) address is refused with the words “Please use your usual email address for the free trial.”, and nothing about it is kept. If you start it, our server keeps on the record for your email when it started and when it ends, where on the site you started it, the label of where your visits came from (see “Where your first visit came from” below; it names our advert if one brought you), the country of your connection at that moment (so the reminder quotes your price), when the reminder was sent, how many videos and clips you finished during the trial, their total length and how many times our AI looked for clips in it (four numbers, never the video or any of its words; counted for a trial started from 3 November 2026, so its last two emails can tell you what you made), when the email on its third day was sent and, if you then start Pro, when you did. If one of our adverts on Facebook or Instagram brought the tab you started it in, the record also holds that advert’s click code and your browser’s name until a Pro payment is credited to the advert or three days after it ends, whichever comes first (see “Our adverts on Facebook and Instagram” below). So that each email and browser gets one trial and each internet connection at most three in 30 days, it also keeps a salted one-way hash of your email address with Gmail’s dots and any +tag removed (until you ask us to delete it), of your browser’s random ID (for a year) and of your connection (for an IPv6 address, of its network part; for 30 days), and of its wider IPv6 network with the trials started on it (for 30 days), each beside the date and a short tag that marks your account (a random code kept on your account’s record; for an account made before 31 October 2026, a scrambled part of its ID). We send it at most four automatic emails through Resend: one when it starts, one on its third day only if you have not made a clip yet and our AI has not looked for clips for you (only for a trial started from 3 November 2026), one the day before it ends and, only if you have not chosen Pro, one when it has ended. The last three each carry a link at the bottom, “Stop these emails”, that stops them in one click, along with the notes when a video is made; sign-in codes still arrive when you ask for one. Someone at Burncaps may also write to you once, in person, to ask how the trial is going; tell them to stop and they will. No card is taken, so Stripe knows nothing of it until you choose Pro. When it ends, your email stays signed in and the account is back on Free (2 videos every 30 days, each with captions and up to 10 clips).
The export check: before an export, the page asks our server for an export pass. It sends a random ID for your browser and a random ID for the file (made up on your device, derived from nothing in the video; your browser keeps it for that file for a day (for a Starter subscription, for 30 days, so that one recording does not count twice), filed under a scrambled form of the file’s name and size that never leaves the device; our own sample talk has a fixed ID instead, so exporting it never uses one of your free videos). It also says whether the export is a whole video or a clip and, for a clip, sends the random ID of the recording it comes from (made the same way); never anything from the video. Free is 2 videos every 30 days, each one recording with its captioned video and up to 10 clips from it. A video counts when its export finishes (the page sends its pass back once, with the same IDs, to say so); a clip counts when its export is allowed, and exporting the same clip again is free. The counts are kept against the browser and against your internet connection, stored as a salted one-way hash of your address (for an IPv6 address, of its network part), with a scrambled form of the random browser ID beside each video counted against the connection, or against your email when signed in to a free account. The random IDs of the clips allowed from a recording are kept with that recording’s count for a day. Each count is kept for 30 days. (From 30 October to 2 November 2026 there were no free videos, and nothing was counted.) When the free videos or a recording’s 10 free clips are used, our server says no, and the page offers Pro and Starter. A Starter subscription asks for a pass for each export, the same way, and our server counts its 6 videos every 30 days (clips from those videos are not counted): it keeps, against a salted one-way hash of your email, the same random file IDs and counts as for a free email account, for the 30-day window, and says no at the limit. A Starter plan paid once before 26 October 2026 exports clips as it was sold; one paid once after that exports no clips. A Pro device, or one on a Starter plan paid once before 8 November 2026, receives a signed pass valid for seven days (during the free Pro trial, until the trial ends), renewed whenever it checks in. The pass carries the same random browser ID, an expiry and what the plan allows (Pro, and whether it exports clips), nothing else. This is what keeps Free, the free trial, Pro and Starter fair for the people who use them. When an export made during the free Pro trial has finished, the page sends the same two IDs once more, with your sign-in, and says whether it was a video or a clip, how long it was, the name of the caption style and the code of the caption language (for example “hi”), so that the note described under “The note when a video is made” below can say what was made; it never carries the video, its file name or its script, and nothing but the length, the style and the language is read from it. If you ticked “Put a small picture of my video in that note” when you signed in at an export (it starts unticked), that message also carries one small still picture of the finished video, with its captions on it (a single frame taken a third of the way in, at most 640 pixels on its long side), only so that note can show it: our server checks it is a small picture, puts it straight into that one email and keeps no copy. No picture is taken while “Keep everything on this device” is ticked.
The note when a video is made (only if you signed in to export): when you sign in to export, and a video or clip you make then finishes, we send one short email, “Your video is made”, to the address you signed in with: how long it was, which caption style and language it used, one small still picture of the finished video (only if you ticked the box for it when you signed in), where your project is kept (in your browser, never on our server) and what Pro adds. We send at most one a day to an account, and only to an account that signed in at an export. Every note has a link at the bottom, “Stop these notes”, that stops them in one click, along with the free Pro trial’s email on its third day, its reminder and its note that the trial has ended; sign-in codes still arrive when you ask for one. Our server keeps, against your account, only that you agreed to the notes at the export, whether you stopped them, and the day the last note was sent (in the free Pro trial, the same report of a finished export also adds one to the trial's count of videos or clips; see the trial below); the video itself and its full script are never sent to us; the one still picture (with the caption words on it at that moment) goes into the email without our server keeping it. Resend, our email provider, keeps a copy of every email it sends, the picture included, for 30 days, then deletes it. The first note an account gets outside the trial also says what stays free. Notes are sent through Resend, our email provider, with no tracking of any kind.
The ready email (only if you ask for it): while a recording is being heard or clipped in your tab, “Email me when ready” in the bar above it sends you one email when the work is done, and only if it took more than two minutes or you had left the tab. Signed in, it goes to your account’s address; not signed in, you type an address, which stays in the page until the work is done, is sent to our server once with that request and is not kept (only a scrambled form of it, which cannot be turned back into the address, is kept to count that a typed address gets at most two a day; an account’s own address gets at most six). The request says whether the captions or the clips are ready, how many clips and how long the recording is; never the video, its file name or its script. If you ticked “Put a small picture of my video in that note” when you signed in, and “Keep everything on this device” is not ticked, it also carries one small still of your video for that email, which our server does not keep; a typed address never gets a picture. The email has nothing in it but the answer to your request. Resend, our email provider, keeps a copy of every email it sends for 30 days, then deletes it. The work happens in your tab, so the email can only come while that tab stays open.
The clip finder, the name check and the caption language: when you choose Clips, type names to spell right, or choose a caption language other than the one spoken, the transcript text of your recording (the words and their timings, and the language they were heard in, never the video or the audio) is sent to our server and passed to Anthropic's Claude API to find clips and rank them (a long recording goes in parts of about eight minutes, then one short list of the best candidates to put them in order), correct the spelling of the names you typed, write a title and a line to post with each clip, or translate the lines into the caption language you chose (for a translation only the lines of text are sent, each word numbered and a mark after the words where the speaker paused, but no times). Anthropic processes API data under its commercial terms and does not train on it. We do not store the transcript (the one exception is a podcast feed you add on Pro, below). The only records our server keeps are counts: a daily count per address, and the number of AI picks used per random browser ID (finding clips uses one for each part of the recording and one to put them in order) (or per email when signed in) and per connection over 30 days (a translation is free and uses none of the picks, only a daily count per address and per random browser ID), each stored as a salted one-way hash, so the feature cannot be abused. With Starter, picks and translations are limited to 230 and 100 in any 30 days: the count is kept per account for each day, and added up over 30 days, only as a safety limit against a misused sign-in. Tick "Keep everything on this device" and this request is never made; a local rule does the job instead.
Step counts: as you use the tool, the page sends our server one count for each step: arriving (with which of our pages you arrived on, for example the home page or a language page), how long the visit stays (once each as the time our pages have been in view in that tab passes 10 seconds, 30 seconds, 2 minutes, 10 minutes and 30 minutes, with which of our pages the visit arrived on; the time stops while the page is out of sight, and the running total is kept in the tab’s own session storage, which your browser clears when the tab closes), dropping or choosing a file on one of our other pages to open it in the editor (with which page, once a visit), using one of our free tools (once a visit for each: a file shown in the safe zone checker, a template saved from it, a subtitle file made or copied in the text to SRT converter, a converted file saved or copied in the subtitle converter, an edited file saved or copied in the SRT editor, a clip saved in the video trimmer, a file saved in the MOV to MP4 converter, a sound file saved in the audio extractor, a louder file saved by the video volume tool, a smaller file saved in the video compressor, a tool’s link to the editor followed, or the subtitle converter’s or the SRT editor’s link to our free SRT generator followed; never anything from the file or the text), starting, the language your recording was heard in (a short code such as "hi" for Hindi, never any of your words) and which speech model wrote it (the larger one, the model made for an Indian language, or our speech service, with the language you picked if the page asked), the caption language you choose and whether its translation came back, whether a video you export uses your own caption look (words appearing as they are said, or your own caption colours; once a visit each, never the colours themselves), each free video made and each video made during the free Pro trial (marked as the trial’s and whether it was the first made on that browser; from 30 October to 2 November 2026 there were no free videos), being told at an export that the free videos or a recording’s free clips are used, or that it needs the trial or a plan (and whether on your browser or on your connection, or, with Starter, for a clip), the reason if you choose to tap one, clicking Starter (and where: the offer under your video, on the card shown when an export needs a plan, or the offer above the script of a recording over ten minutes), clicking to start Starter at $11.98 a month (and where: the Starter page, the Pro page, the home page or the card shown when an export needs a plan), clicking to start Pro (and where: the Pro page, the Starter page, the home page, the offer under your video, the Pro offer beside your clips, the card for a recording over ten minutes or the card shown when an export needs a plan), coming back to our page after one of our buy buttons opened a checkout, when no payment followed (once for each press, with which plan button it was, where it was on the site and how long you were away, in four bands: under 10 seconds, 10 to 60 seconds, 1 to 5 minutes, or over 5 minutes; the time of the press is kept only in the tab’s own session storage, and whether a payment followed is read only from this browser’s own record of your plan), your answer if you tap one to the question then shown beside that button, “What stopped you?” (the answer’s short code, the plan button and where it was; there is nothing to type), starting the free Pro trial (and where: the card shown when an export needs a plan, the offer under your video, the plans card, the Pro offer beside your clips, the home page, the first screen shown to visitors from our adverts, the Starter page, the Pro page or the sign-in when you export), which step of starting the free Pro trial was reached (its card shown, its button pressed, an email typed for the code, the code sent or not sent, a new code asked for, a code that did not match, the code accepted, the Google sign-in accepted, the trial refused, or the plans shown to a sign-in that has no plan; never the email or the code), the next step chosen under a finished video while the free Pro trial is on (finding the clips in that video, going back to its clips, or choosing a longer recording), the sign-in asked for when you press Export without being signed in (whether it was shown, shown with a free Pro trial on offer, skipped with “Export without signing in”, or finished; never an email), asking for the ready email (whether you asked, whether it went, or the work finished too soon to need it; never the address), your script and result appearing on screen (once a visit), pressing Export (once a visit), a stretch of your recording that had sound but came back with no words (once a file, and whether listening to it again found them), a file turned away for being too big or too long (with its length rounded up to a band of minutes, such as up to 2, up to 15 or up to 30, or that the browser could not say), a file your browser stopped being able to read, or that the video player refused, on the way to your script (with the stage it happened at and nothing else: while the video was being opened, while its sound was being read, while the rest of its sound was being read, after an export had started, or that the player refused it, with the browser’s own number for the refusal; never the file, its name, its size or anything in it), a recording over ten minutes opened on a computer for its script and subtitle files (once a file, with its length in the same bands), the clips of a recording found (once for each script, with how many were found, in bands such as 4 to 9, or that the simple rule on your device chose them, and then why in a word: you kept everything on your device, the recording was too short, the AI found none or could not be reached, or a limit was used), opening a clip (once a visit for each place in the list: the top three, 4 to 10, or 11 to 20), pressing Export all (once a visit, with how many clips were chosen, in bands), pressing Add B-roll on a clip (once a visit for each answer: how many moments it found, none to five, or why it found none, in a word: busy, paused, a limit used or failed), changing a B-roll shot (once a visit for each kind of change: swapped, moved, removed, or all removed; never the words, the search words or the shot), opening the Pro offer beside the clips of a recording (once a visit), leaving your email on a “Tell me when it opens” card (and where the card was: the home page, the Starter page, the Pro page or under your video), leaving your email for our emails (and where: under your video), sending yourself this page from a phone, whether an export or the clip finder finished or failed (and, for a failed export, the stage it stopped at and why in a word, such as the video decoder refused the file, stopped answering, or the file could no longer be read; when the video decoder or player failed, the browser’s own short words for the fault, the video format’s code, how many frames had gone in and come out, and the format’s number for the kind of frame it started on, never anything of the picture or the sound), whether our speech service could be used for a recording and, when it could not, why in a word (it was busy, not working, paused, or a limit was used), a count if the editor could not load in your browser at all, and a count the first time in a visit you play with Wick, our mascot, in each way (a poke, a drag, feeding it a video, taking its offer to try the sample, its party, hiding it or bringing it back), a count the first time in a visit you try the sample video (a musician’s talk to camera, with its clips already found) from the button under the drop box or from the link in one of our emails that opens it (and which of the two), and, on a phone inside an app’s own browser (Instagram’s, for example), whether our note about opening the page in Chrome or Safari was shown beside your finished video and whether you tapped to open it or copied the link (until 31 October 2026 it was shown above the drop box too, with a choice to stay). On a phone the page also counts each stage as it starts (the file chooser opened, or closed without a file, and, the first time it closed without a file, that a short help appeared under the drop box and which of its choices you tapped: choose again, send this page to your computer, how to get your own video from YouTube, or the sample video, or that you closed it; the file chosen, with its size in megabytes, its width and height in pixels, its length in seconds and the phone's memory size; the speech model loading; listening; finished; and for an export, its size and length, whether a background was on, which video encoder the phone used, how long finding the face took, how far it had got each quarter and each minute, how many B-roll stock shots it read and whether one was left out because it could not be fetched or took over 15 seconds, and at the end how many frames a second it wrote and how many milliseconds each stage of a frame took on average; whether the page went out of sight while it listened or wrote a video, how far the video had got and how many seconds the page was away; the memory the page’s own code was using, where the browser says (Chrome does), at the start of an export and at each quarter; the stage at which a failure the page told you about happened, and why in a word; when an attempt was started again because the page had gone out of sight, under a lower label for the video’s format, or on our own software decoder, which runs in your browser; when the sound of a video was read by our own software sound reader, which also runs in your browser; what stopped a file from opening, that the browser could no longer read a file (with how many seconds after you chose it, whether it was the first time for that video, and whether the page had been out of sight since), and how a long whole video was written and saved (exp-sink-disk, exp-sink-blob, saved-big or saved-big-back), in a word (for example that it was empty, or which kind of file it is inside); when the page sent the sound to our speech service again after a dropped connection, and whether it then got through; when pieces already heard for the same file were used again instead of being sent; when the part of a video already written was kept on the phone, whether a later export carried on from it and from how far (a percentage, and for Export all how many of its clips were already saved), and whether it finished; and, when the sound could not be read in pieces, why in a word), because a phone can close a page without warning and the last count shows where it stopped. Each count carries only the kind of place your visit came from (for example ChatGPT, Google, Bing, a social site, or none), the kind of place your first ever visit came from and roughly how long ago that was (the same day, the next day, within a week, or longer: see “Where your first visit came from” below), whether you are on a phone or a computer, whether the page is open inside an app’s own browser (Instagram’s, Facebook’s, Messenger’s, TikTok’s or another app’s, read from the name your browser gives itself, never which account), and the country your connection is in, which our host reads from the request (your internet address is not kept). In the counts our server keeps there is no identifier of any kind and nothing links one count to another, so they are plain totals, not a record of you; the main steps among them also go to PostHog, under a coded visit ID that changes every day (How the site is used). They tell us where people come from and where the tool stops working for them.
Where your first visit came from: the first time you open Burncaps, your browser keeps a small note in its own storage on your device: the kind of place that visit came from (as for the step counts), our advert’s tag or a partner’s code if the link carried one, and the date, with no time. It leaves your device only as a short label (for example “org-chatgpt”) on the step counts, on a payment and when you start the free Pro trial, so we can see which places bring people who come back and buy. Clearing this site’s data in your browser removes it. A browser that used Burncaps before 2 October 2026 is labelled “earlier”, because its first visit is not known.
A plan before it opens: if you leave your email on one of the “Tell me when it opens” cards, our server keeps the email, which card, where on the site the card was, the label of where your first visit came from (above), the country of your connection and the date. We use it for one email, when that product opens, and for nothing else; ask us to delete it at any time.
Emails from Burncaps (only if you ask): after an export you can leave your email for our emails. Our server then keeps the email, where on the site you left it, the label of where your first visit came from (above), the country of your connection, the date, whether you have confirmed it, and which of our emails it has been sent. First we send one email with a link to confirm the address is yours (and at most one more, if you leave it again a day or more later). Until you confirm, we send it nothing else, and if you do not confirm within seven days of the last of those emails, we delete the address and everything above. Once you confirm, you get one email when something big arrives and, unless you have a paid plan, three short emails over the next ten days on getting more from Burncaps (what the paid plans change, how buyers use them, and the price against a year of monthly plans). Every email has a link at the bottom that stops them in one click; your mail app’s own Unsubscribe button works too. They are sent through Resend, our email provider, with no tracking of any kind: no pixel, and nothing that follows one person. A link back to burncaps.app carries only a label for which email it was (for example src=email-n1), the same for everyone who gets that email, so we can count which emails bring people back. We keep the email until you stop them or ask us to delete it; after a stop we keep only a note that the address stopped, filed under a scrambled form of it (a salted one-way hash), so you are not written to again; the email itself, and everything else above, is deleted.
Emailing yourself this page (phones): on a phone you can send the link to burncaps.app to your own email, to open on a computer. The address goes to our server and on to Resend for that one email, and is not kept: we count how many links each address and each connection asked for that day, stored as salted one-way hashes, so nobody can use the button to send mail to strangers.
Problem reports: when the page meets an error it cannot recover from, or a step fails and the page has to try another way, it sends our server the error message, the place in our code where it happened, the page and your browser’s name, never your video, its sound or its transcript. The “Report a problem” link sends the words you type, the page, your browser’s name and, only if you leave it in the box, your email, so we can reply (a signed-in page fills in your account’s email; you can clear it). When the box still holds the email of the account you are signed in with, the report also carries your sign-in, so our server can tell a customer’s report from an anonymous one; with the report it keeps only whether it came from a signed-in or a paying account. A report you write is emailed to the founder the moment it arrives, up to a few an hour from one connection or one account. When it carries your email, it is answered within 30 minutes: your words, the page, your browser’s name and, when you are signed in, your plan go to Anthropic’s Claude, which drafts an answer to exactly what you wrote; the founder can send it, change it or write his own, and if he has not within 25 minutes, the draft is sent to you from hello@burncaps.app. A push to the founder’s phone (through ntfy.sh) says only what kind of note arrived, never your words or your email. We keep the note, the answer and whether it was sent with the report. Reports are kept in our fix queue: up to 20 written reports a day from one connection, 10 from one signed-in account (30 from a paying one), and up to 30 of the page’s own automatic reports a day from one connection; past that, the page says the report was not sent. To keep these limits, our server counts reports by the hour and by the day under a scrambled form of the connection or the account (a salted one-way hash), and deletes those counts after two days. Mail you send to hello@burncaps.app reaches the founder’s inbox, and a copy of its sender, subject and text (never its attachments) goes into the same fix queue, so a problem you email is seen as fast as one you report here.
If we wrote to you about listing Burncaps: we sometimes write, one person at a time, to people who publish lists, reviews or resource pages about video, caption, subtitle or podcast tools, using the work address or the contact form on their own site. We keep your address, the subject, the date and what we sent in our records, so we never write to the same address twice within 30 days. Reply “stop” or “no” and we never write again. The email is plain text with no tracking of any kind: no pixel, no rewritten links and nothing that follows one person. It comes from our own address, hello@burncaps.app, sent either by hand from that mailbox or by our server through Resend, our email service. Ask us to delete it all at any time: the address, our records and the note go (and with them our record that you asked us to stop).
If we emailed you as an agency or production company: from 2 November 2026 we write to companies that make clips or edit podcasts and talking videos for clients: firms in the United States and limited companies in the United Kingdom only. We use the business address the company publishes on its own website or, since 9 November 2026, the work address and job title of its owner or a senior person from SuperSearch, Instantly’s database of business contacts. We keep the address, the person’s name and job title where we have them, the company’s name, website and country, its company number (UK), the page we found it on, one sentence about its work taken from that page, and the emails sent and any reply. They are held in our list on this computer and in Instantly, the service that sends these emails from mailboxes on the domains entrysummit.co, keepertrade.co, causewaycommerce.co and driftcommerce.co. Before our first email we check that the address can receive mail, with an email-checking service run on Apify; only the address is sent to it. At most two emails go to one company, in plain text with no open or link tracking. Reply “no” and we never write again: we keep only the address on a do-not-contact list so it stays that way. If you do not reply, we delete your details 90 days after our last email. Ask us to delete everything at any time.
If we emailed you about a clip from your show: from 4 November 2026 we write to people who publish long video episodes (podcasts and interview shows) in the United States, Canada, Australia, New Zealand and Brazil (in Portuguese). We use the contact address your podcast’s public feed or your own website gives. Some of these emails say we have already made a clip: to show you what Burncaps does, we made one clip from an episode you posted publicly, with Burncaps on our own computer. Others offer to make clips from an episode you send us, and we make them only if you reply with one. A clip we make is yours: we never post it, never use it on our site or in adverts and never show it to anyone else, and we send it only if you ask. We keep your address, your name and your show’s name, the episode’s address, where we found your address, and the emails sent and any reply, in our list on this computer and in Instantly (the same mailboxes as above). Before our first email we check that the address can receive mail, with an email-checking service run on Apify; only the address is sent to it. At most two emails go to one show, in plain text with no open or link tracking. Reply “no” and we never write again: we keep only the address on a do-not-contact list so it stays that way. We delete our copy of the episode and the clip 30 days after our last email, and your details 90 days after it if you do not reply. Ask us to delete everything at any time.
The price where you are: Starter is $11.98 a month, £11.98 in the UK, €11.98 in the euro area and ₹998 in India, and US dollars everywhere else, or by the year where a yearly price is offered; Pro’s price depends on the country. When a page opens, it asks our server which of these applies. The request carries nothing about you: our host reads the country from it, as for the step counts, and nothing is kept. The answer stays in your browser’s session memory for that tab only.
How the site is used (PostHog): since 7 November 2026 we use PostHog, an analytics service (PostHog Inc., on its EU cloud in Frankfurt, Germany, under its data processing agreement with us), to see how people use Burncaps so we can make it better: which page a visit arrives on, how far it gets and where it stops. Only the main step counts above also go to PostHog, with the same details and labels: arriving, how long the visit stays, starting, pressing Export, each export or clip finished or failed, a file your browser could no longer read, a page reopened on a phone after an earlier try stopped, a file turned away for being too big or too long, our speech service or a caption translation failing, reaching a limit of Free and the reason if you tap one, each step of starting the free Pro trial, clicking a plan’s button, and coming back from a checkout with no payment and your answer to the question then shown. Every other step count goes only to our server. PostHog also gets repeated quick clicks on one spot, a sign that something did not respond, sent as the kind of element and its place in the page, never its words or where it leads. Page views, how far down a page you scrolled and other clicks are not sent to PostHog. Each event also carries the page’s address with only our own tags kept (for example utm_source: anything else after the ? is removed, and anything shaped like an email is replaced), the page that linked to it, your browser’s name as it gives it, its language and time zone, your operating system, your screen’s size, the country of your connection and its price group (which our server reads from the request, as for the step counts), and, if you are signed in, your plan’s name and the day of your free Pro trial. PostHog runs only on burncaps.app, starts only once a page has finished loading, and reaches PostHog through our own address (burncaps.app/ingest), which passes it on. Nothing is stored on your device for it: no cookie and nothing in your browser’s storage; what it needs is kept in the page’s memory and is gone when the page closes. To tell one visit from another, each page asks our server for a coded visit ID (/api/visit): a code our server makes from your internet address and your browser’s name with a random key made each day (from midnight UTC) and deleted when the next day’s is made, so the code changes every day, and once the key is gone nobody, us included, can tell which address it came from. Our server keeps neither your address nor the code, and PostHog is set to discard internet addresses. If you are signed in (a free account, the free Pro trial, Starter or Pro), the page sends your sign-in with that request and your events are filed under a coded account ID instead: a code our server makes from your account, the same on every device and every day, so we can see whether a trial is used across its 7 days; it is never your email, and PostHog cannot turn it back into your email. Our server also tells PostHog, under that coded account ID, when your free Pro trial starts (with where on the site and the country) and when it ends (with how many videos and clips were made in it), when an export starts, is refused (with why, in a word) or finishes, whether it is a whole video or a clip, and when a plan is bought, cancelled, ends or starts again, by the plan’s name only: never an amount, a price or your email. Our own browsers and our own accounts are never sent.
Recordings of the page (outside the European Economic Area): PostHog records how a page looked and changed from the moment one of these happens on it: pressing a plan’s button, coming back from a checkout with no payment or answering the question then shown, tapping a reason at a limit of Free, reaching a limit of Free other than the one for a long recording’s whole video, an export or a clip failing, a file your browser could no longer read, or starting the free Pro trial. For about one page in twenty it also records from the moment of one of these: reaching the limit for a long recording’s whole video, a file turned away for being too big or too long, a page reopened on a phone after an earlier try stopped, a step of starting the free Pro trial, or pressing Export; and about one page in twenty opened while you are signed in is recorded from the start. No other page is recorded. A recording shows our own words, buttons and layout, and where the mouse or finger moved, tapped and scrolled, so we can see where people get stuck. It is rebuilt from the page’s code, not filmed from your screen. Never recorded, and shown as an empty box: your video and every frame of it, the canvases the captions are drawn on, sound, every picture made on your device or shown in the editor, and Wick. Hidden: everything you type, and every word or label the page did not have when it was published, such as your transcript, your clips’ titles and captions, your file names and your email, which show only as asterisks. No recording is made of the page you come back to after a payment or of a page whose address carries anything other than our own tags, and a recording pauses while the page listens to your sound or writes your video. In the European Economic Area, or where the country is unknown, no page is recorded; the events above are still sent. PostHog keeps the events for up to a year and the recordings for 30 days.
Turning PostHog off: Turn PostHog off in this browser (your browser then keeps one small note, bc_ph_off, only to remember that choice) or turn it back on. It is also off whenever your browser sends Global Privacy Control or Do Not Track, and while “Keep everything on this device” is ticked in the editor. We do this on the basis of our legitimate interest in understanding how our site is used so we can improve it. In the UK this needs no consent because it is used only for that, you are told about it here, and you can object at any time, free, with the link above (the Privacy and Electronic Communications Regulations, regulation 6 and Schedule A1, as amended by the Data (Use and Access) Act 2025).
Posting to YouTube (only when you press Post to YouTube)
Post to YouTube is waiting for Google’s review and is not on the site yet. This is how it will work. After you export a video or a clip you can post it to your own YouTube channel. None of this happens until you press Post to YouTube.
- Signing in. Your browser opens Google’s sign-in window (accounts.google.com). You choose the account with your channel and allow one permission: to upload videos (Google’s screen calls it “Manage your YouTube videos”). We ask for nothing else: not your email address, your name or the videos already on your channel.
- Where the permission goes. Google sends the permission (an access token) to our page burncaps.app/youtube-connect.html, in the part of the address that is never sent to any server; that page passes it, sealed, to your Burncaps tab and closes. Burncaps keeps the token in that tab’s memory only: we never save it on your device, never send it to our server and never log it. Your browser does keep the address Google returned to, with the token in it, in its own history (and in its sync, if you use that) until the token expires, within the hour. Closing the tab, or “Forget the YouTube sign-in on this tab”, makes Burncaps forget it; to end it at Google at once, remove Burncaps as below.
- The video goes straight to YouTube. The file and the details you typed (the title, the description, its hashtags as tags, who can watch it, the time you scheduled, and your answer to “Is this made for kids?”) go from your browser straight to YouTube (www.googleapis.com). Our server never receives the video, the details or the token, and we keep no record of what you posted.
- YouTube API Services. Burncaps uses YouTube API Services to post for you. What YouTube does with your video is set by the YouTube Terms of Service and the Google Privacy Policy. Through YouTube API Services we access none of your YouTube data; we use your permission only to upload the file you chose with the details you typed; we store none of it and share it with no one. Burncaps’ use of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements.
- Taking it back. You can remove Burncaps’ permission at any time in your Google Account’s security settings: myaccount.google.com/connections (or security.google.com/settings/security/permissions). That ends the token at once, including the copy in your browser’s history.
- What the page counts. One step count,
yt-post, when a post finishes or stops, with one word for how (posted, private, scheduled, quota, signed out, refused, connection, stopped): never the title, the file, the account or the channel. It joins the step counts above on the day Post to YouTube opens.
What we do not do
- No cookies.
- No advertising, no data sales.
- No AI training on what you make. Your video never reaches us, so we cannot keep it or train on it, and we train no AI model on your sound, your words or anything else you make. For the speech service, Microsoft says its fast transcription “does not use customer data to improve its Speech models”; Microsoft lists MAI-Transcribe-2, the model it runs for us, as a preview, and its pages do not yet say whether this covers previews. For the words sent to Claude, Anthropic’s commercial terms say it “may not train models on Customer Content”. What thirteen video editors’ own terms say about your videos and AI: AI training in video editors, compared.
- No account needed for Free. A record is made only for a Starter or Pro purchase, when you sign in with your email (to export, or because your connection's free videos were used), or when you start the free Pro trial or a payment you try fails, and it holds only what is listed above.
- No tracking for advertising and no following you across other websites. The one report about a person to an advertising company is the one to Meta described under “Our adverts on Facebook and Instagram”, made only after a tap on one of our adverts: that the page opened, your browser’s first video, and a trial start or a payment. How we see the way our own site is used is under How the site is used (PostHog): a visit is known only by a code that changes every day, unless you are signed in.
Visit counting
We use Cloudflare Web Analytics, a cookieless page-view counter, to see how many people visit and which pages they read. It sets no cookies, does not fingerprint your device, and cannot identify you. It is the one small script from cloudflareinsights.com you will see in your browser's network tab alongside the AI model download. See Cloudflare's Web Analytics documentation for what they collect. PostHog, described under How the site is used, counts visits too; its requests go to burncaps.app/ingest.
Our own adverts
A link from one of our own adverts carries a short source tag (for example src=rd-anchor) so we can count which advert brought visitors, how many of them tried the tool, and how many bought a plan. The tag is held in your browser's session memory for that tab only and is gone when you close it; if it brought your first ever visit, it is also part of the note described under “Where your first visit came from”. It is not a cookie, it contains nothing about you, and the counts are plain totals with no identifiers. A link from one of our partners carries a short code (ref=); your browser keeps the code and the date for 60 days so a purchase in that time can be credited to the partner, and nothing else. If you arrive without a tag, the step counts above are recorded as "organic". When you choose a plan, the label of where your first visit came from goes to Stripe with the payment and, if this visit came from somewhere else (an advert, a partner), that label too (for example “org-chatgpt” or “org-chatgpt_ms-cards”), so a sale can be credited to what brought it; it says nothing about you. No advertising network's tracking code runs on this site.
Our adverts on Facebook and Instagram
When you tap one of our adverts on Facebook or Instagram, the link that opens Burncaps carries Facebook’s click code (fbclid) beside our own advert tag. Your browser keeps the click code in its session memory for that tab only, and it is gone when you close the tab. While it holds the click code, our server tells Meta, the company behind Facebook and Instagram, about these moments, so that Meta can count what the advert brought: that the page the advert opened has loaded (once per tab); that the first video ever started in your browser has begun, if you start it in that tab (once per browser, never for later videos); that the first video ever made in your browser is finished, if you make it in that tab (once per browser, never for later videos); and, if in that tab you start the free Pro trial or pay for Starter or Pro, the trial or the sale. For the page and the first video, your browser sends our server the click code and the page’s path, and our server sends Meta the click code, the internet address and the browser name of that request, the page’s address without anything after the path, the time and a random ID so that it is never counted twice: no email and nothing from your video. For a trial or a payment, the request that starts the trial or switches your plan on also carries the click code, and our server sends the click code, your email in hashed form, the internet address and the browser name of that request, the page, the time, an ID for the trial or the payment so that it is never counted twice, and for a payment its value and currency. If you start the free Pro trial from such a tab, our server keeps the click code and your browser’s name with the trial until three days after it ends, so that a Pro payment at its end, made in any tab, can be credited to the advert: when that first payment arrives, our server tells Meta once, with the click code, your browser’s name, your email in hashed form and the payment’s value and currency (not your internet address), and deletes the click code. If you do not choose Pro, the click code is deleted then. Hashing turns your email into a fixed code of 64 letters and digits that does not spell out the address: the same email always gives the same code, so Meta can only match it to an account that already has that email. Your email itself is never sent to Meta. If you came any other way, including a link to us in a Facebook or Instagram post, profile or message, nothing is sent to Meta. No Facebook script, pixel or cookie is on our pages. Tick “Keep everything on this device” in the editor and the click code is not kept, so nothing is sent. We do this on the basis of our legitimate interest in knowing which of our adverts bring customers, so that we pay only for the ones that do; you can object at any time by ticking that box or by writing to hello@burncaps.app, and we stop.
Hosting
The site's static files are served by Netlify, which may log IP addresses of requests for security and operations, as any web host does.
Your rights
Under UK GDPR you have rights over personal data we hold. Because we hold almost none (at most, your Starter or Pro sign-in record described above, a Pro podcast feed and its episodes, an email you left on a “Tell me when it opens” card, or an email you left for our emails), a single email to hello@burncaps.app is enough to exercise any of them. PostHog holds the events and recordings of a signed-in account under its coded account ID; ask us and we delete them there too.