Whisper API

WhisperAPI is a hosted transcription service for audio and video files, built on the open source Whisper speech recognition model. Files are sent through an HTTP API or a drag and drop dashboard, and transcripts come back in minutes in JSON, plain text, VTT, DOCX or PDF, among other formats. The vendor says it runs its own self hosted copy of Whisper, so files are not sent to OpenAI or another third party AI provider. Developers can choose between Whisper model sizes to trade speed against accuracy, upload files directly or pass remote URLs, tune model parameters, and handle both audio and video through the same endpoint. Uploads can be as large as 10GB, there is no minute limit, and more than 98 languages are supported. Uploaded files are deleted automatically after 24 hours, and only the transcript text stays in the account. The vendor states that it never uses customer content to train or improve any model. It suits developers adding transcription to a product, and non developers who want a simple dashboard. Billing is pay as you go through prepaid credits that do not expire, with no monthly fee. The credit cost of a job depends on model size, audio length billed per started hour, speaker diarization and file size. Bundles are 40 credits for $10, 100 credits for $20, 200 credits for $30, and custom purchases from 1,000 credits at $0.10 per credit. New accounts receive 5 free credits, and custom enterprise pricing is available on request.