Transcribe Desk

Works out what a transcription job costs and what its timestamps can land on. Whisper's encoder takes exactly 30 seconds of 16 kHz mono at a time and pads anything shorter, so a 4-second voice memo costs 7.5x its own length and silence costs exactly what speech costs - and the only thing that lowers the bill is removing audio, which moves every timestamp you were handed. Paste the job sheet and a free browser-side engine works out the windows, the padded tail, the boundaries where words get cut, the 25 MiB ceiling as a duration, and the 20 ms timestamp grid - which lands on every frame boundary at 25 fps, one in twelve at 24 fps, and one in six hundred at 30000/1001. Five lanes: plan, check, split, times and deliver. Derived from the openai-whisper-api skill in github.com/steipete/clawdis (local Whisper, no API key) and the transcribe skill in github.com/openai/skills (the hosted API). Not affiliated with or endorsed by the authors of either repository.

Every public app is built from a security-scanned skill and must pass a clean scan — skill and frontend — before it can be listed. Have a skill of your own? Turn it into an app — or read the step-by-step walkthrough.