ScribeToAny
Turn any audio or video into accurate text
ScribeToAny turns audio and video files into text and subtitles. It transcribes about 98 languages with timestamps and speaker labels, lets you edit every line in the browser, exports TXT, SRT, VTT, TSV, CSV, JSON, PDF and DOCX, and translates the result into 134+ languages with subtitle timings kept. You can start free with 100 minutes every month, no card needed.
Replies usually arrive within 1–2 business days.
Who builds it
I'm Ray Mac, a former ByteDance engineer. I build and run ScribeToAny on my own; the first version went live in July 2026. The tested guides on the blog are mine too: I run what they describe on real recordings and files before publishing.
English isn't my first language, and a lot of what I watch and work with isn't in English either. After leaving ByteDance, I kept running into the same problem: tools for turning recordings into text and subtitles were English-first, meeting-first and quick to hit you with limits. So I built the tool I wanted: about 98 languages in, 134+ out, subtitles checked for reading speed, and a free allowance that renews every month instead of a one-time trial.
ScribeToAny is operated by Scribe Saas, LLC, a company registered in the United States.
How it works
Uploads go straight from your browser to encrypted storage. Transcription runs Whisper-class speech models (faster-whisper) on dedicated GPUs, detects the spoken language automatically and labels speakers. Subtitles are checked for reading speed: cues that go by too fast are flagged, never silently rewritten.
Engineering notes: running Whisper on GPUs from Cloudflare WorkersHow it's tested
I publish what I measure, including where it falls short:
- Accuracy: on a scripted test recording, 4 of 189 words differed from the script (a 2.1% word error rate), and all 14 speaker turns were labelled correctly. Read the details
- Subtitle files: running 16 broken SRT files through ffmpeg and our own parser turned up three bugs in the parser, which are now fixed. Read the details
- Languages: when Thai, Lao and Burmese recordings came back with speech missing, I traced it to two decoding settings and changed them for those scripts.
- Pricing: the comparison with Otter, Rev, Sonix and HappyScribe also says where they beat ScribeToAny. Read the details
Your files
- The original audio or video is deleted automatically after transcription. If a job fails, the file is kept for up to 7 days so you can retry.
- Transcripts stay in your account until you delete them.
- We never use anything you upload to train AI models.