Turning a recording into text usually means uploading it to a server, creating an account, and paying per minute once a short free trial runs out. This mp3 to text converter does the job differently: drop in an MP3, WAV, M4A or video file and OpenAI Whisper transcribes it right in your browser, with no upload, no sign-up and no per-minute fee.
Everything runs on your own device, so your recording is never sent to a server. That is what makes this safe for private recordings like interviews, lectures, meetings and voice memos. The transcript comes with timed segments, so the same page doubles as a subtitle generator: download the result as plain text, or as SRT or VTT subtitles ready for a video editor.
Key features
- Whisper transcription in your browser: the same open source speech model behind most commercial transcription services, running locally through Transformers.js, with no account and no API key.
- Two models to choose from: Fast (Whisper tiny) for quick results on phones and older laptops, or Balanced (Whisper base) for clearer reading of unclear speech. The model downloads once on the first run and is then cached.
- Timed segments for subtitles: every part of the transcript carries a start and end time, so SRT and VTT downloads drop straight into a video project.
- Audio and video files: MP3, WAV, M4A, AAC, OGG, FLAC and Opus audio, plus MP4, MOV and WebM video, whose soundtrack is transcribed. Files up to 100 MB and recordings up to 30 minutes are accepted.
- Private by design: your recording is decoded and transcribed on your own device and is never uploaded, so there is no server copy and no upload wait.
- Free, with no sign-up: no account, no credits and no per-minute limit.
How to use
- Drop your audio or video file onto the tool, or choose it from your device. Files up to 100 MB are accepted.
- Pick Fast for a quick transcript or Balanced for better accuracy on unclear speech. The first run downloads the model once, about 40 MB for Fast or about 75 MB for Balanced, and your browser caches it afterwards.
- Click Transcribe and leave the tab open while the work runs on your device. A long file takes a while, and slower devices take longer.
- Read the transcript with its timed segments below the tool, copy it, or download it as TXT, SRT or VTT.
Why Whisper, and why in the browser
Whisper is OpenAI’s open source speech recognition model, trained on 680,000 hours of audio. Commercial services built on it charge per minute and require your recording to be uploaded to their servers. Running the small Whisper models in the browser removes both costs: there is no server bill to pass on, and a recording you could not send to a third party never has to leave your device. The trade off is speed and size: only the tiny and base models fit comfortably in a browser, so this tool is built for short to medium recordings, not hour long files on a low end phone.
Private by design
Like every Savozi tool, this one runs entirely in your browser. Your recording is decoded and transcribed on your own device and is never uploaded to our server, so there is no server copy and no upload wait. It is part of our watermark tools collection of free photo, video and audio tools: free, no account, and nothing to install.
Honest limits
This version transcribes English speech only, because it uses the English only Whisper tiny and base models. Recordings in other languages will transcribe poorly or not at all. The tiny model is fast but makes more mistakes than the base model, and both struggle with heavy accents, overlapping speakers, music and noisy recordings, so proofread the result before quoting or publishing it. Timestamps mark timed segments, not individual words, and there are no speaker labels. Video files only work when your browser can decode their audio track, so a video with an unusual audio codec may fail to decode: try an MP3 or WAV version of the same recording in that case.