Transcribe Service

Self-hosted speech-to-text via faster-whisper. Streams progress over SSE.

Version: 1.0.0

Demo

Drop a video or audio file below to transcribe it live.

Drop video or audio here
or click to choose a file

No file selected.

Memory

1,690 MBUsed
13,877 MBAvailable (system)
11,829 MBAvailable for jobs
15,568 MBTotal
0 MBReserved (0 jobs)
2,048 MBOS reserve

POST /transcribe

Returns text/event-stream. JSON body with gcs_url or multipart upload at POST /transcribe/upload.

{
  "gcs_url": "https://storage.googleapis.com/bucket/video.mp4",
  "quality": "fast"
}

Quality presets

qualitymodelRAM / job
fastsmall int8~2.5 GB
balancedmedium int8~4 GB
bestlarge-v3-turbo int8~7 GB

SSE events

event: started
event: segment
event: done
event: error

Auth

Optional — currently open (no AUTH_TOKEN set).

GET /health

JSON status, loaded models, active jobs, memory stats.