Unlimited local audio & video transcription. Your audio never leaves your machine.
Podcasters pay monthly for hosting, transcription and audiograms forever, even for a back catalog that no longer grows. Pay-once audio tools cap the cost of a show at a single payment.
The tool this job usually means renting is Otter.ai, at $16.99/mo — roughly $204 a year, every year. WhisperDesk does the job for a one-time $39 — running locally on your own machine.
mp3, wav, m4a, mp4, mov, mkv, flac, ogg, webm — video audio is extracted automatically.
For podcasters, drag & drop anything pays off for every episode, including the back catalog — and it's covered by the same one-time $39, not metered per month.
Transcription runs entirely on your machine. Nothing is ever uploaded.
For podcasters, 100% private pays off for every episode, including the back catalog — and it's covered by the same one-time $39, not metered per month.
No cap, no monthly quota. Transcribe 10 hours a day if you want.
For podcasters, unlimited minutes pays off for every episode, including the back catalog — and it's covered by the same one-time $39, not metered per month.
One-time $39 for the packaged Windows installer — no terminal, no build step.
On first run WhisperDesk fetches whisper.cpp (~8 MB) and your chosen model (78–148+ MB) with a visible progress bar. One time only.
Timestamped text in minutes, exportable as TXT, SRT or VTT — all offline.
WhisperDesk is $39, once. Otter.ai at $16.99/mo — roughly $204 a year, every year. Pays for itself in under 3 months of Otter.ai Pro.
Pays for itself in under 3 months of Otter.ai Pro. WhisperDesk costs $39 once; Otter.ai charges $16.99/mo — about $204 a year. For podcasters, that fixed cost is the whole point: the tool is a one-time line item, not a monthly one.
Yes — MIT-licensed at github.com/bensblueprints/whisper-transcriber-mvp, and it always will be. The $39 buys the signed installer, 1-click setup and updates.
It runs OpenAI's Whisper models — the same family behind most modern transcription tools. Bigger models (small/medium) are more accurate and slower; you pick the trade-off.