AI Transcription

Yes. 25 free transcriptions per day, no signup needed. Maestra and Otter both gate their best features behind paid plans; here you get OpenAI's Whisper-quality output for free up to your daily cap. Pro users get unlimited transcriptions.

MP3, WAV, M4A, WebM, OGG, FLAC for audio; MP4, MOV, MKV, AVI for video. Maximum file size 25 MB (Whisper API limit). For longer recordings, split into chunks or compress the audio.

Auto-detect plus 18 manually selectable languages: English, Spanish, French, German, Italian, Portuguese, Russian, Japanese, Chinese, Korean, Arabic, Hindi, Urdu, Turkish, Dutch, Polish, Indonesian, Vietnamese. Whisper supports ~100 languages internally; ours covers the highest-traffic locales.

Five: SRT (subtitles with HH:MM:SS,mmm), VTT (web subtitles with dot separator), TXT (plain text), Markdown (timestamp headers + speakers), JSON (full structured transcript). One-click copy or download for each.

Whisper-1's basic model does NOT include speaker diarization (who said what). The transcript groups text by acoustic boundaries, not voices. For dedicated speaker labeling, consider Maestra's paid tier or AssemblyAI. We may add diarization via a different provider in a future release.

OpenAI Whisper achieves around 95% word accuracy on clear audio in English and 85-90% across major languages, matching or exceeding Otter, Rev, and Descript. Accuracy drops for: heavy accents, fast speech, background noise, music. For best results: clear single-speaker audio, 44.1kHz or higher, no music.

Uploaded files go to OpenAI's Whisper API for processing and are not retained by Molixa. Per OpenAI's API terms, your audio is not used to train their models. The transcript is returned to your browser; we do not store it on our servers beyond the rate-limit usage log (no transcript content stored).

That's OpenAI Whisper's hard cap per request. Roughly: 25 MB equals 25-30 minutes of compressed MP3 or ~10 minutes of WAV. To transcribe longer content, split into chunks first (we recommend ffmpeg) or compress to 64kbps MP3.

The result viewer shows segments with timestamps. Copy the text and paste into any editor. We may add an inline editor with click-to-seek playback in a future release. For full timeline editing with synced playback, Descript or Otter are stronger choices today.

Maestra is the closest peer: both use Whisper-class models. Their advantage: voice cloning, live translation, dubbing (we don't have these). Our advantage: free unlimited use of core transcription, no signup, modern export pipeline, integration with Molixa's translator and summarizer for downstream workflows.