Whisper as a server, with your own API
Audio transcription via OpenAI-compatible API
- Video memory
- 8 GB
- Setup
- ~4 min
- Access
- port 8000
- Billing
- hourly, in BRL
Whisper transcribes audio accurately, including noisy recordings and strong accents. Running on your machine, you pay for the GPU hour and transcribe as many hours of audio as fit in it — there is no per-minute charge.
faster-whisper-server exposes Whisper large-v3 via 100% OpenAI-compatible API (/v1/audio/transcriptions). High-quality transcription with diarization support.
What it is for
- Automatic meeting minutes, with timestamps
- Batch subtitling for an entire video catalog
- Support-call transcription for quality review
- Sensitive audio processed inside your own machine
How to deploy
- Create your account and add balance (card or Pix, no subscription).
- In the console, pick the Whisper Server template and a machine — the console hides the ones that do not meet the requirement.
- In about 4 minutes the setup finishes and the access address shows up in the panel, on port 8000.
Done? Just destroy the machine and billing stops with it. No contract, no minimum commitment.
FAQ
How many hours of audio per GPU hour?
With the large model on an 8–16 GB card, transcription runs several times faster than real time, which usually means dozens of audio hours per rented hour.
Does it identify speakers?
Plain Whisper transcribes without separating speakers. Add a diarisation step to your pipeline for that.
Is there a compatible API?
Yes, the server exposes an HTTP API you can call from n8n, a script or your own backend.