WAAS

WAAS offers a REST API with endpoints for transcription, language detection, and job management. The API supports queuing, so multiple jobs can be processed asynchronously. Users can choose from different Whisper models (e.g., tiny) and specify tasks like transcribe or translate. Results can be downloaded in various formats including SRT, VTT, JSON, and plain text. The GUI, Jojo, provides an editor that runs locally in the browser, allowing users to listen to segments and correct transcription errors. Webhook support is also available for automated notifications.

WAAS can be used in various scenarios where audio or video transcription is needed. For example, media companies like VG (a Norwegian publisher) use it to transcribe interviews or broadcasts. Developers can integrate the API into their applications to add speech-to-text capabilities. The queuing system makes it suitable for processing large volumes of files. The editor feature is useful for post-processing transcriptions to improve accuracy. Additionally, the webhook functionality enables automated workflows, such as triggering downstream processes after transcription is complete.

WAAS is an open-source project, so the software itself is free to use and modify. However, running it requires infrastructure, including a machine with sufficient VRAM for the Whisper model (the smallest model requires about 1GB of VRAM). Users are responsible for hosting and maintaining their own instance. There is no mention of any paid service or pricing plans on the page.