How to auto-generate subtitles from a video
- Drop the file. MP4, MOV, WebM, MP3, WAV or M4A. The browser reads the audio track locally first; nothing is sent anywhere yet.
- Press “Generate subtitles”. The audio (up to 10 minutes, about 25 MB once encoded) is sent to Vercel AI Gateway, which routes it to a speech recognition model and returns timed text.
- Fix what the AI got wrong. Click any line to jump the video there, correct words, drag timings with “Start here” and “End here”, split or merge lines. Lines that flash by too fast are flagged in amber.
- Export. Download an .srt for YouTube, Premiere, DaVinci Resolve or VLC, a .vtt for web players, a plain transcript, or render a new video with the captions drawn into the picture.
How the transcription works
Many online subtitle generators send your whole video to a server, queue it, and email or display the result later. SubtitleWave only ever uploads the audio track, once, when you press “Generate subtitles”, and only to Vercel AI Gateway, which routes the request to a speech recognition model built for timed transcripts. The audio is used to produce the transcript and is not stored afterwards. Files are capped at 10 minutes and about 25 MB so the request stays fast and affordable to run for free.
The optional AI text tools (translation, punctuation clean-up, shortening fast lines, YouTube chapters) go through the same Gateway but only ever receive the subtitle text you choose to send, never the audio or video.
How accurate is it?
Accuracy depends mostly on the audio: one clear voice close to the microphone comes out nearly word-perfect; music beds, crosstalk and distant speakers need more hand editing. Names, brands and jargon are the usual misses, which is why every line stays editable and a find-and-replace sits one click away.
Speed: what to expect
Generating subtitles uploads the audio and waits for the model's response, so timing depends mostly on your connection and clip length: a few seconds for a short clip, roughly a minute or so for a full 10-minute file. Subtitles appear on the timeline once the transcript comes back, ready to edit.
Questions people ask
Is this subtitle generator really free?
Yes, within a fair-use limit per visitor so the service stays free for everyone: no sign-up, no watermark, up to 10 minutes of audio per file.
Is my video uploaded?
Only the audio track, and only once you press “Generate subtitles”. It is sent to Vercel AI Gateway to be transcribed and is not stored afterwards. The optional AI text tools send subtitle text only.
Which languages can it transcribe?
The speech recognition model handles a wide range of spoken languages. Leave “Detect automatically” on, or pick the language to improve accuracy. The AI translator can also convert finished subtitles into 36 other languages.
Can I use the SRT file on YouTube, TikTok or Premiere?
Yes. SRT is accepted by YouTube Studio, Facebook, LinkedIn, Vimeo, Premiere Pro, Final Cut Pro (via import), DaVinci Resolve, CapCut desktop and VLC. TikTok and Instagram do not accept caption files, so for those use the burned-in video export.
Is there a length limit?
Yes: up to 10 minutes of audio and about 25 MB once encoded, per file. Split longer recordings into parts and generate subtitles for each.
Does it work on a phone?
Yes: your phone only has to play the file and upload its audio, so it works well even on modest phones.