Generate subtitles for any video or audio
AI speech-to-text that runs entirely in your browser. Auto-detects the language and exports SRT, VTT, or a plain transcript.
Free · No signup · Files never leave your device
Drag and drop a video or audio file
or click to browse
How it works
Drop your video or audio
Any MP4, MOV, MKV, MP3, WAV, or podcast recording — it stays on your device.
AI transcribes locally
OpenAI's Whisper speech-recognition model runs right in your browser via WebGPU or WebAssembly. The model downloads once — your file is never uploaded.
Download your subtitles
Get a timestamped SRT or VTT subtitle file or a plain-text transcript — free, no watermark. Pick how much text each caption holds, down to word-by-word.
Subtitles without uploading your footage
Every online subtitle generator asks you to upload your video to their servers and wait in a queue. This one works differently: OpenAI's Whisper speech-recognition model downloads to your browser once (cached for next time), and every file after that is transcribed on your own hardware. Nothing is uploaded, ever — which matters for client work, internal meetings, interviews, unreleased content, or anything under NDA. On a machine with GPU acceleration it transcribes several times faster than real time.
The output is a standard SRT file with numbered, timestamped segments — the format YouTube, LinkedIn, Instagram, Premiere Pro, DaVinci Resolve, CapCut, and every media player accept directly. Choose VTT instead for HTML5 web video, or plain text when you just need the transcript for show notes, meeting minutes, or a blog post. Formats and caption lengths switch instantly after transcription, and the full text copies to your clipboard in one click. For platforms that ignore subtitle files, the same tool can burn styled captions permanently into the video itself, or add a soft subtitle track viewers can toggle — both run in your browser too. Already have a subtitle file? Upload your own SRT or VTT and skip transcription entirely.
Tips for accurate captions
Clear speech with modest background noise transcribes best. If you know the spoken language, picking it from the dropdown instead of auto-detect removes one source of error — especially for short clips. For noisy recordings, running the file through the free noise remover first noticeably improves results, and the video trimmer lets you cut a long recording down to just the part you need captions for.
Frequently asked questions
Yes. The subtitle generator is completely free, with no signup, watermarks, or usage limits. We make money from our video translation service, not from this tool.
No. The Whisper AI model runs locally in your browser using WebGPU or WebAssembly. Your video or audio never leaves your device — once the model has loaded, the tool even works offline.
Whisper was trained on about 100 languages, including English, Spanish, French, German, Chinese, Japanese, Arabic, and Hindi. Language is auto-detected by default; picking the spoken language from the dropdown can improve accuracy.
SRT (the universal subtitle format accepted by YouTube, Premiere, DaVinci Resolve, and every video player), VTT (for web video and HTML5 players), or a plain-text transcript. You can switch formats after transcribing without re-processing.
Yes — the same tool does it. After transcribing (or uploading your own SRT/VTT) a video, switch to "Burn into video" to bake styled captions permanently into the picture (TikTok-style word-by-word text, background box, custom size and position), or "Soft track" to add a subtitle track viewers can toggle on and off.
Yes. If you already have an SRT or VTT file, upload it instead of transcribing — then download it in another format, burn the captions into the video, or add them as a soft track.
The Whisper speech model (~77–150 MB depending on your browser) downloads the first time you use the tool and is cached by your browser. After that, transcription starts immediately.
Whisper is one of the most accurate open speech-recognition models available, and clear speech transcribes very well. Heavy background music, strong accents, or crosstalk lower accuracy — the timestamped output is easy to touch up in any subtitle editor.