Steps
Generate subtitles from the speech
Drop the video (MP4, WebM, MOV and most audio files work). Choose the accuracy: Balanced suits most clear speech, Best handles accents, noise and other languages better but is slower. The first run downloads the model once (the size is shown before you start: about 150 MB for Fast, 235 MB for Balanced and 355 MB for Best). Pick the spoken language or leave automatic detection on, press Transcribe, then download the .srt file. Whisper does not label speakers and can mishear names, numbers and jargon.
Correct the text and timing
Open the .srt file and the video. Play it through and fix misheard words, names and numbers. Cues that overlap, have lines over 42 characters or go by too fast to read are highlighted. Snap a cue to the picture with Set start and Set end; if every subtitle is early or late, use Shift, and if they drift over time, use two-point sync. Press Sort & fix, then download .srt (or .vtt).
Burn them in, or upload the file
For Instagram, TikTok, WhatsApp, presentations or anywhere caption files are not supported, burn the subtitles into the picture: choose the video and the corrected .srt, pick a style, size and position, check the preview and press Burn in subtitles. In current browsers this runs several times faster than the video plays; where a browser cannot encode video directly, it falls back to recording in real time, so keep the tab visible. Burned-in subtitles cannot be turned off or edited later, so keep the original video and the .srt. For YouTube, you can skip this step and upload the .srt as a caption file instead.
Requirements and sources
| What | Value | Source |
|---|---|---|
| YouTube - SubRip (.srt) caption files | Only basic versions of these files are supported. No style info (markup) is recognized. The file must be in plain UTF-8. | Supported subtitle and closed caption files - YouTube Help Last verified 2026-10-21 |
| YouTube - WebVTT (.vtt) caption files | In initial implementation. Positioning is supported, but styling is limited to <b>, <i>, <u> | Supported subtitle and closed caption files - YouTube Help Last verified 2026-10-21 |
| WCAG 2.2 Success Criterion 1.2.2 Captions (Prerecorded), Level A | Captions are provided for all prerecorded audio content in synchronized media, except when the media is a media alternative for text and is clearly labeled as such. | Understanding Success Criterion 1.2.2: Captions (Prerecorded) - W3C WAI Last verified 2026-10-21 |
| WebVTT format (W3C Candidate Recommendation Draft, 20 May 2026) | This specification defines WebVTT, the Web Video Text Tracks format. | WebVTT: The Web Video Text Tracks Format - W3C Last verified 2026-10-21 |
FAQ
Should I burn in subtitles or upload an SRT file?
Upload the file where the platform supports caption files, such as YouTube: viewers can switch it on or off and you can correct it later. Burn them in for places that ignore or limit caption files, or when you want a fixed style. Many people do both: burn a copy for social media and keep the .srt for YouTube.
How accurate are the automatic subtitles?
Clear speech from one person at a time works best. Accents, music, crosstalk and background noise lower accuracy, and names, brand names and numbers are the most common errors. Whisper's accuracy also varies by language. Always watch the whole video in the editor before publishing.
Is my video uploaded to make the subtitles?
No. The Whisper model is downloaded from Hugging Face and cached by your browser; after that, your video is decoded and transcribed on your device. The editor and the burn-in step also work locally.
Can I subtitle a video in another language, or translate it to English?
Choose the spoken language from the list (it covers 22 common languages plus automatic detection) and use Best for languages other than English. To get English subtitles from foreign-language speech, set Output to Translate to English. This option translates into English only.
Does this work on a phone?
Yes, but it is slower than on a computer, and the model download is large, so use Wi-Fi. Use Fast or Balanced on older phones. Keep the screen on and the tab in front while it works.
Why does WCAG matter for my videos?
WCAG 2.2 success criterion 1.2.2 (Level A) asks for captions on prerecorded audio in synchronized media. Organisations that follow WCAG, such as many public bodies, schools and companies, expect captions on their videos. Edited captions are more reliable for this than unchecked automatic ones.
BufferMint is not affiliated with YouTube, Google, Instagram, TikTok or OpenAI. Platform caption rules are set by each service and change; check their help pages. Files are processed in your browser and are not uploaded.