LansonAI
Console

The Voice Context Layer for Live Speech.

Turn live speech into stable, readable, multilingual context. Real-time and recorded speech APIs with explicit contracts for transcript stability, latency, and translation.

Start

5-minute quickstart

quickstart.py
from lansonai import LansonClient

client = LansonClient(api_key="ls_live_...")

with client.realtime(language="en") as session:
    session.send_audio(microphone_stream)
    for event in session.events():
        if event.type == "transcript.stable":
            print(event.text)

What do you want to build?

Choose your use case — each path has a dedicated guide and API reference.

    Real-time captions

    Stream audio over WebSocket and receive transcript events with explicit partial, stabilized, stable, and final states.

    Live translation

    Source text and translation arrive as paired segments. Switch languages mid-session.

    Recorded audio

    Upload a file or URL and get a transcript, word-level timestamps, speaker diarization, and SRT/VTT subtitles.

Core concepts

Understand the model before you integrate.

    Voice Context Layer

    What it is, and how it differs from raw STT output.

    Stable vs. Partial Text

    Why realtime text is not a token stream, and when text is safe to render.

    Latency Model

    Network, recognition, stabilization, and translation — where time goes.

    Trinity Engine

    The shared technology foundation behind LansonAI's speech processing.

For coding agents

This documentation is designed to be human-readable, agent-readable, and schema-accurate.

    llms.txt

    Machine-readable documentation index for coding agents.

    OpenAPI spec

    Schema-accurate API specification.

    Swagger UI

    Interactive API explorer for REST endpoints.

Ready to integrate live speech?

Follow the 5-minute quickstart, then pick Realtime or Recorded depending on whether you need streaming or file transcription.
Get started5-minute quickstart