Turn French audio into text in your browser: interviews, lectures, meetings, voice memos or the sound of a video, as a plain transcript or as timed .srt subtitles. French is hard for speech to text in a particular way. Many endings are silent, so parle, parles and parlent sound the same, and the model has to choose the spelling from the rest of the sentence. On clear speech Whisper gets most of them right; agreements and homophones are where to look when you proofread. The recording never leaves your machine.
How it works
- Drop a French audio file or video. French is already selected, so the decoder writes French from the first word.
- Choose Accurate (≈77 MB) for Québécois, Belgian, Swiss or African accents and for noisy rooms; Fast (≈41 MB) for a quick pass on clear speech. Switch Output to Subtitles before running if you need .srt.
- Press Transcribe, then proofread the silent endings: play a line, correct it in place, and download .txt, .srt or .vtt.
Questions
- Why do some verb endings and accents come out wrong?
- Because they sound identical. Il parle and ils parlent, parlé and parler, a and à are pronounced alike, so the spelling has to come from context. In a short or cut-off phrase there is too little context and the small models guess. Edit the line in place; the timing stays as it was.
- Can it translate French audio to English text?
- Not in one step. This tool writes down what was said, in French. For an English version, run it with Output set to Subtitles, download the .srt and open it in the subtitle translator: in Chrome 138 or later it translates each line on your device and keeps the timings. Correct the French first, because translation carries every transcription error with it.
- Does it handle Québécois and African French?
- Yes, under the same French setting, which covers every variety. Standard broadcast French gives the cleanest transcript; strong regional accents, fast casual speech and local vocabulary produce more errors, and place names and personal names are the most common misses on any accent. The Accurate model copes better, so use it for anything you plan to publish.
- Does my file get uploaded?
- No. The models and all processing run inside your browser on your own device. You can verify this yourself: open your browser’s DevTools, watch the Network tab, and process a file: there are no uploads. After the first visit most tools also work fully offline.
- How accurate is it?
- The models are OpenAI’s Whisper (tiny and base tiers). On clear speech they are very usable; heavy accents, crosstalk and noisy rooms will show errors. The Accurate tier is noticeably better on difficult audio.
- Which languages work?
- Fifteen, picked in the Language setting: English, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Polish, Turkish, Japanese, Korean, Chinese, Arabic and Hindi. The decoder is conditioned on the one you pick, so it transcribes that language rather than approximating it in English — which also means picking the wrong one returns confident nonsense. The weights understand 99 languages; these are the ones we measured as good enough to offer.
- How long does it take?
- Transcription runs on your processor, so speed depends on the machine: expect around a quarter of the recording’s length, and less on the Fast model than on Accurate.