Whisper

Whisper (opens in a new tab) is an automatic speech recognition system that enables transcription in multiple languages.

We provide the mosec inference service template.

Mosec

You can deploy the pre-built template here (opens in a new tab) to deploy the service. Run the inference with SDK.

export MODELZ_API_KEY=mzi-abcdefg...
modelz inference --deployment whisper-demo-XXX --serde raw --read-stdin <speech_16000.mp3

Python Client

import modelz
 
 
APIKey = "mzi-abcdefg..."
 
cli = modelz.ModelzClient(deployment="whisper-demo-XXX", key=APIKey)
with open("speech_16000.mp3", "rb") as f:
    data = f.read()
resp = cli.inference(params=data, serde="raw")
print(resp.data)

Stable Diffusion Streamlit