The Captions tab does two things and charges nothing for either: it detects the speech in a video you upload, and it burns caption lines into the frames with local ffmpeg. Both run on our own hardware. The part that can cost money is the transcription step, and only sometimes — this page is straight about when, because it is the number nobody else spells out.
Upload a video, press detect, and the speech comes back as timed lines you can edit before anything is rendered. Pick a style, a colour and a position, and the lines are burned into the frames with ffmpeg on our own server. The output is a new video file with the captions baked in — not a sidecar subtitle file, which is worth knowing if you wanted an .srt.
The editing step in the middle is the point. Automatic transcription is good and not perfect: proper nouns, brand names and overlapping speech are where it slips. Fixing four words before the burn costs nothing and is faster than rendering twice.
Which plan: The Audio Studio opens on the Starter plan. That gate lives on the server: every generating audio endpoint answers 403 to a free account, so it is not a greyed-out button you can work around. Everything behind these tabs runs on a paid provider, which means it is billed in ✦ gold credits — ⚡ green credits, the ones that refill daily, cannot pay for any of it.
| Style | What it looks like |
|---|---|
| Cinematic | Large, bold, letter-spaced and uppercase, with a soft fade in and out |
| Subtitle | Clean lower-third captions, the conventional broadcast look |
| Karaoke | Bold and punchy, with a short fade as each line lands |
| Typewriter | Monospaced, thin outline, no shadow |
| Neon Glow | Heavy blurred outline, for vertical social cuts |
| Minimal | Small, light, no outline and no shadow at all |
Speech-to-text on this platform runs on our own self-hosted Whisper service, and while that service is answering it costs nothing. Pressing a microphone should not have a price on it. But it is a ladder, not a guarantee: when the local container is restarting or the GPU is busy, the request steps onto OpenAI's Whisper API, and past that onto a third provider.
Watch out: When transcription falls through to a paid provider it bills ✦1 gold per started minute of audio — and a free account holds no gold, so it cannot pay that. The burn-in itself stays free either way. Most of the time the local service answers and nothing is charged; the point is that free speech-to-text is true for the common case, not for every case.
A short mic press bills the one-credit minimum when it lands on the paid rung. A ten-minute video that has fallen through costs ✦10 for the transcription pass. Nothing is charged when a transcription fails: the balance check happens before the provider call and the deduction only after it returns.
A separate subtitle file is the better format in exactly one place: a player you control, where a viewer can switch captions on. Everywhere a short clip actually gets watched — a feed that autoplays muted, a repost, a file somebody drops into an editor — the subtitle track is either not read at all or left switched off. Burning the lines into the frames is the version that survives the trip.
The trade is that the text is permanent, and sized for the frame it was rendered into. If a clip is going out both wide and vertical, caption each crop separately rather than captioning once and re-cropping afterwards: a bottom-positioned line on a 16:9 master ends up outside a 9:16 cut.
Timing comes from the speech, so the lines are already in sync when they arrive. Editing the words does not move the timing, which is what you want when you are fixing a misheard name — but it also means a line you rewrite much longer stays on screen only as long as the phrase it replaced.
The second mode has nothing to do with speech. Type a line, upload a clip, and the text is burned across the whole video in whichever of the six styles you picked. It is the fastest way to put a title, a handle or a call to action onto a vertical cut, and because it never touches transcription it is free with no conditions attached.
| Limit | The number | Note |
|---|---|---|
| Caption lines a video | 500 | Refused above this, with a message |
| Video size for the transcription step | 200 MB | Trim or compress first |
| Burn-in cost | 0 credits | Local ffmpeg, not a provider |
| Transcription cost | 0, or ✦1 a started minute | Free on our own Whisper; billed when it falls through |
| Plan | Starter and up | Enforced at the route |
One consequence of the tab living inside the Audio Studio: the caption burn-in is free but not reachable free. A free account is refused at the endpoint along with everything else behind that gate, so the 0-credit price is a price for paid accounts.
Open the Captions tab — Opens the studio on Captions. Upload, detect, edit and burn all happen there.
The burn-in is 0 credits — it is local ffmpeg. The transcription that produces the lines is free while our own Whisper service answers, and bills ✦1 gold per started minute when it falls through to a paid provider.
Yes, and you should. Every detected line is editable in place, which is where names and numbers get fixed. Up to 500 lines are accepted per video.
No. Captions are burned into the frames and you get back a new video file. That is what social platforms and autoplay feeds need; it also means the text cannot be switched off later.
The transcription step takes files up to 200 MB. Beyond that, trim or compress the clip first — the interface refuses it with a message rather than failing part way through.
No. The Captions tab sits inside the Audio Studio, which is Starter and up and enforced by the server. The 0-credit burn-in price applies to accounts that can reach it.
Yes — Static Text mode does exactly that, in any of the six styles, and it never touches transcription, so it is free with no conditions.