TurboScribe & Whisper
The English-side transcription stack: OpenAI's open-source Whisper model, the apps built on it, and TurboScribe for bulk files.
On this page
Whisper is OpenAI's speech-recognition model. It is open source, so dozens of apps run it, and it is the engine behind much of what members use for English audio. For shiurim that mix Hebrew, Yiddish and Aramaic, Sofer.ai still does better; the comparison is on Transcription.
TurboScribe
TurboScribe is the group's standard answer for "I have a pile of recordings to transcribe." Members say it runs on Whisper, and they have repeated the same line since 2025: three free transcriptions a day, a flat price for unlimited, and one member "transcribed hundreds of hours in a single month."
Plans as listed on TurboScribe's pricing page in October 2026 (the site blocks automated checks, so confirm before you pay):
| Plan | Price | Limits |
|---|---|---|
| Free | $0 | 3 files a day, each up to 30 minutes, one upload at a time, slower queue |
| Unlimited | $20/month, or $10/month billed yearly ($120) | Files up to 10 hours or 5 GB, 50 uploads at a time |
What members say: good accuracy on English and French, works for WhatsApp audio files, speaker labels "not the best job" (May 2026). In August 2026 a rabbi's advice for a Q&A article from an interview was TurboScribe first, then an AI pass to clean it up.
Whisper itself
- Open source, MIT license (github.com/openai/whisper). Free to run on your own computer if you are comfortable installing it, or ask Claude Code to set it up for you; a shliach did that in May 2026 to transcribe and rename a library of old Gemara shiur recordings by masechta and daf.
- Through an API, for agents and automations. OpenAI's current transcription models cost $0.18 to $0.36 per hour of audio, and Groq hosts Whisper for $0.04 to $0.11 per hour. Full price table on Transcription.
- Hebrew and Yiddish versions. The ivrit.ai project publishes Whisper models tuned for Hebrew and for Yiddish on Hugging Face, and runs the free transcribe.ivrit.ai service for Hebrew (Google sign-in required).
Apps built on Whisper
- MacWhisper: runs Whisper on your Mac, recommended for heavy users (November 2025). Check its page for the current price.
- Synopsule: records and transcribes on-device with Whisper on Mac and iPhone, with an Apple Watch recorder; a $1.99 one-time purchase covers all three. Shared in June 2026 for its privacy angle.
- A home-built tool. In August 2026 a rabbi had Claude build him a local transcription app that takes YouTube links, streams, uploads and device audio, with several Whisper-class models and a queue. He still uses Sofer.ai for English full of Yiddish terms.
Other free options members use
- Google Pinpoint (journaliststudio.google.com/pinpoint): Google's research tool for journalists, which transcribes audio and video. Members say files can be up to 2 hours with no limit on the number of files (January 2026).
- Google Recorder on Pixel phones: members say running the transcription a second time improves it.
- iPhone Voice Memos and Samsung's recorder app make transcripts you can paste into ChatGPT (June 2026).
- Adobe Podcast Enhance: a free filter that cleans up a bad recording before you transcribe it. Members called it "fantastic" for class recordings.
- Letterly: a dictation app that tidies rambling speech into text; a member said it can shape a voice ramble into a drasha outline (February 2026).
NVIDIA Parakeet and Canary
Free, fast speech-recognition models that came up in the group in 2025. NVIDIA's model cards list English only for Parakeet v2, 25 European languages for Parakeet v3, and English, German, French and Spanish for Canary 180m Flash. None lists Hebrew or Yiddish. They are a good choice for English in a local app (Handy offers Parakeet), not for a Hebrew shiur. A member found Parakeet "a bit slow" on his machine in July 2026.
Workflow
- Record (phone voice memo, Zoom, lapel mic).
- Pick the tool by language: Sofer.ai for mixed Torah audio, transcribe.ivrit.ai for Hebrew, TurboScribe or another Whisper tool for English.
- Transcribe, then fix names and Hebrew terms.
- Hand the transcript to Claude for a summary, a shiur outline or action items.