Documentation

Everything you need to integrate Sayd into your application.

Speech-to-Text

Real-time streaming transcription with < 200ms latency. Supports batch and streaming modes.

Real-time Streaming

WebSocket-based audio streaming for live transcription with word-level timestamps.

Talk API

Push-to-talk voice input with AI-powered transcript cleaning. Send audio, get polished text.

Quick Start

Install the SDK and make your first API call:

$ pip install sayd

Python

import sayd

client = sayd.Client(api_key="sk-your-key")

# Transcribe an audio file
result = client.stt.transcribe(
    model="sayd-v1",
    audio=open("audio.wav", "rb"),
    language="auto",
)
print(result.text)

# Stream audio for real-time transcription
for chunk in client.stt.stream(audio_stream, model="sayd-v1"):
    print(chunk.text, end="", flush=True)

Authentication

All API requests require an API key. Get yours from the dashboard and include it in requests:

cURL

# Include your API key in the Authorization header
curl -X POST https://api.sayd.dev/v1/stt/transcribe \
  -H "Authorization: Bearer sk-your-key" \
  -F model="sayd-v1" \
  -F language="auto" \
  -F audio=@audio.wav

Talk API — Voice Input with AI Cleaning

The Talk API combines real-time speech-to-text with LLM-powered transcript cleaning. Send raw audio via WebSocket, get polished, publication-ready text back. Perfect for push-to-talk interfaces, voice notes, and dictation features.

Python

import sayd

client = sayd.Client(api_key="sk-your-key")

# Create a Talk session for real-time voice input with AI cleaning
session = client.talk.create(
    language="auto",       # "en", "zh", or "multi"
    sample_rate=16000,     # 8000 or 16000 Hz
    codec="pcm16",         # "pcm16" or "opus"
    cleaning_level="standard",  # "light", "standard", "aggressive"
    output_format="paragraph",  # "paragraph", "bullets", "raw"
)

# Connect to the WebSocket and stream audio
ws = session.connect()

# Stream PCM16 audio frames
for chunk in audio_source:
    ws.send_bytes(chunk)

# Signal end of recording — the server will automatically
# drain any in-flight audio frames (up to 500ms), wait for
# STT to stabilize, then run LLM cleaning.
ws.send_json({"type": "end"})

# Receive the AI-cleaned transcript
result = ws.receive()  # {"type": "cleaned", "text": "..."}
print(result["text"])

WebSocket Protocol

After creating a session via POST /v1/talk, connect to the returned WebSocket URL to stream audio and receive results.

Bash

# WebSocket Protocol — Talk API

## Connect
ws = WebSocket("wss://api.sayd.dev/v1/talk/stream/{session_id}?api_key=sk-...")

## Server Messages (you receive)
{"type": "ready"}                    # Session ready, start sending audio
{"type": "partial", "text": "..."}   # Interim transcript (may change)
{"type": "sentence", "segments": []} # Final transcript segment
{"type": "cleaned", "text": "..."}   # ✨ AI-cleaned result
{"type": "complete", ...}            # Session complete

## Client Messages (you send)
[binary PCM16 frames]               # Raw audio data
{"type": "end"}                      # Signal end of recording
{"type": "keepalive"}                # Keep connection alive

## End Signal Behavior
# After receiving "end", the server continues accepting
# in-flight audio for up to 500ms (drain window), ensuring
# no trailing speech is lost. Then it waits for STT to
# stabilize and runs LLM cleaning. No client-side delay needed.

API Endpoints

POST/v1/stt/transcribeTranscribe audio file

WS/v1/stt/streamStream audio for real-time transcription

POST/v1/talkCreate a Talk session (returns WebSocket URL)

WS/v1/talk/stream/{id}Stream audio with AI cleaning

GET/v1/talkList Talk sessions

GET/v1/talk/{id}Get Talk session details & results

SDK

Python

Official Python SDK — supports both sync and async clients.

pip install sayd