01
Your app
Product, workflow tool, agent system, or internal service.
Izwi Server
A local API for private speech workflows
Point your application at Izwi’s OpenAI-compatible endpoints. Keep a familiar request shape while speech processing runs close to the data.
Local server request
from openai import OpenAI client = OpenAI( base_url="http://localhost:8080/v1", api_key="not-needed",) audio = client.audio.transcriptions.create( model="qwen3-asr-0.6b", file=open("call.wav", "rb"),)Architecture
A small, inspectable integration path
The integration model is simple: application, request, runtime, environment.
02
/v1 request
A familiar API shape for transcription and speech workflows.
03
Izwi Server
The runtime process that coordinates local models and audio work.
04
Private environment
Laptop, workstation, server, VPC, or controlled deployment target.
Use server for
Integration when the workflow is ready for product code
Use Desktop to evaluate models and audio. Use Server when an application is ready to call the workflow through an API.
- add private transcription to an application
- keep audio processing behind your infrastructure boundary
- reuse OpenAI-compatible client patterns
- prepare for supported enterprise deployment when production requirements are clear
Wire private voice into your application
Start with the documentation for a local integration. Talk to an engineer when you need a supported deployment in a defined production environment.