Free Transcription — Audio & Video to Text

The first 30 minutes are genuinely free — no credit card, no trial countdown, full features (speaker labels, 99 languages, all five export formats including SRT). Here's the honest part most free-tool pages skip: it's a one-time 30 minutes, not a monthly allowance. If you have a single lecture, interview, or voicemail to transcribe, free covers it completely. If you transcribe regularly, paid plans start at $2/month for 200 minutes — the cheapest entry tier in the category as of July 2026.

Every 'free' transcription tool has a limit — the difference is how visibly they state it. Common catches: file size caps (Canva's converter caps at 4.5 MB, roughly 4 minutes of MP3), per-file duration caps (TurboScribe and UniScribe cap free files at 30 minutes), monthly minute pools (Otter Basic gives 300 min/month but caps each recording at 30 minutes and limits file imports), watermarks or export restrictions (several tools give you the transcript on-screen but charge for the download), and account-required 'free' tiers designed as an upsell funnel. The only genuinely unlimited free option is self-hosted Whisper — because you provide the computer.

Yes, one honest way: self-hosted OpenAI Whisper. It's the same model family behind most commercial AI transcription (VexaScribe included), open-source and free forever. Requirements: a computer with Python installed, ideally a GPU (a 1-hour file takes 10-20 minutes on an RTX 3060-class GPU vs 30 minutes to 2 hours on CPU), and comfort with the command line. Install with pip install openai-whisper, then run whisper yourfile.mp3 --model large-v3. No speaker labels out of the box — add pyannote.audio or use WhisperX for that. If that setup sounds like a project, the cloud free tiers listed on this page cover occasional use without it.

Speaker labels (diarization) are the feature most commonly cut from free tiers. VexaScribe includes full speaker diarization in the free 30 minutes — same quality as paid, no gating. Otter Basic includes speaker identification on its free tier for live meeting recordings. Most converter-style free tools (Canva, Zamzar-class sites) don't offer speaker labels at all, free or paid. Self-hosted: base Whisper doesn't do speakers — you'd need WhisperX or pyannote.audio 3.1, which is a genuinely technical setup.

A few tools offer no-signup transcription (audiototext.com, NoteGPT claim this), with the trade-off that you can't save, edit, or re-export the transcript later — and quality/limits vary. VexaScribe requires an account for the free 30 minutes because the transcript lives in an editor you can return to, with speaker renaming and five export formats. If your priority is absolute zero-friction for a short, disposable transcript, a no-signup tool works. If you want the transcript to be editable and exportable, the 30-second signup is the trade.

For rough personal reference, often yes. For anything you'll share or publish, usually no. YouTube auto-captions run roughly 82-92% accuracy on clean English — meaning 8-18 errors per 100 words — with no punctuation reliability and no speaker labels. The workflow is also awkward: upload your file to YouTube (unlisted), wait for caption processing, then extract the captions via YouTube Studio. It's genuinely free and unlimited, which is why we list it, but the accuracy gap vs Whisper-class transcription (91-97%) is noticeable on anything that matters.

Free tier accuracy equals paid accuracy on most cloud tools — the free tier limits quantity, not quality. VexaScribe's free 30 minutes run the same Whisper Large-v3 model as paid plans: 95-97% on clean single-speaker audio, 91-95% on Zoom-quality recordings, lower on noisy or heavily accented audio. Where free options genuinely lose accuracy: YouTube auto-captions (82-92%), older free tools running smaller models, and any tool that processes only a low-bitrate compressed version of your upload. The bigger hidden cost of weak free tools is proofreading time — a 90% accurate transcript of a 1-hour file needs 30+ minutes of correction.

Same tools, same limits — video is just audio with pictures attached. VexaScribe's free 30 minutes accept video directly (MP4, MOV, MKV, WebM and more, up to 5 GB) and extract the audio automatically. One free-tier gotcha specific to video: file-size-capped tools hit their limits much faster with video (a 4.5 MB cap fits ~4 minutes of MP3 but only ~30 seconds of 1080p video). If your source is video and the tool caps file size, extract the audio first with a free tool, or use a duration-capped tool instead.