Saidwire Speech-to-Text

Community-maintained integration. This service is built and maintained by SaidwireLLC. Pipecat does not test or officially support it. Please report issues and request changes on the source repository. Learn more about community integrations.

Overview

SaidwireSTTService provides real-time English speech-to-text through Saidwire. It extends Pipecat’s WebsocketSTTService, so connection management, reconnection, keepalive and metrics follow the same patterns as the built-in services. Interim results are pushed as InterimTranscriptionFrame and finals as TranscriptionFrame. When a VADUserStoppedSpeakingFrame arrives, the service sends Finalize, and Saidwire answers right away with the final transcript of everything heard so far. That final is pushed with finalized=True, so turn-taking doesn’t wait on a server-side pause.

Source Repository

Source code, README, changelog and issues

Example

A voice agent that hears the user through Saidwire

Saidwire API Docs

Streaming protocol, parameters, limits and error codes

PyPI Package

The pipecat-saidwire package on PyPI

Installation

This is a community-maintained package distributed separately from pipecat-ai:

uv add pipecat-saidwire

Prerequisites

Saidwire Account Setup

Before using the Saidwire STT service, you need:

Saidwire Account: Sign up at saidwire.com/dev
API Key: Create a key from your dashboard

Required Environment Variables

SAIDWIRE_API_KEY: Your Saidwire API key

Configuration

str

default:"None"

Saidwire API key. Defaults to the SAIDWIRE_API_KEY environment variable.

str

default:"wss://api.saidwire.com/v1/listen"

Streaming endpoint.

int

default:"None"

Sample rate of the incoming audio in Hz (8000 to 48000). If None, uses the pipeline’s input sample rate.

bool

default:"True"

Send Finalize when a VADUserStoppedSpeakingFrame arrives, so the final transcript comes back right away.

float

default:"300.0"

Seconds to keep retrying while the service answers “starting up” or “busy” (HTTP 503 or 429) before reporting a permanent error.

SaidwireSTTService.Settings

default:"None"

Runtime-updatable settings. See Settings below.

float

default:"0.30"

P99 latency from speech end to final transcript, in seconds, used by turn strategies. Measure your own deployment with stt-benchmark and pass the result.

Settings

Runtime-configurable settings passed via the settings constructor argument using SaidwireSTTService.Settings(...). They can be updated mid-conversation with STTUpdateSettingsFrame; a change reconnects the stream between user turns.

Parameter	Type	Default	Description
`model`	`str`	`"saidwire-en-2026-10-30"`	Model id.
`language`	`str`	`"en"`	`"en"`, `"en-US"`, `"en-GB"`, … (English only).
`endpointing`	`int \| bool \| None`	`None`	Pause in ms that ends an utterance on the server. `False` leaves turns to your VAD.
`interim_results`	`bool`	`True`	Interim transcripts while the user speaks.
`punctuate`	`bool`	`True`	Punctuation and capitals.
`early_words`	`bool`	`False`	Fire `on_early_word` for each word as soon as it is committed.
`vad_events`	`bool`	`False`	Fire `on_speech_started` when the server hears speech begin.
`utterance_end_ms`	`int \| None`	`None`	Fire `on_utterance_end` after this much silence.

Usage

Basic Setup

import os

from pipecat.pipeline.pipeline import Pipeline
from pipecat_saidwire import SaidwireSTTService

stt = SaidwireSTTService(api_key=os.getenv("SAIDWIRE_API_KEY"))

pipeline = Pipeline(
    [
        transport.input(),
        stt,
        user_aggregator,
        llm,
        tts,
        transport.output(),
        assistant_aggregator,
    ]
)

With Early Words

from pipecat_saidwire import SaidwireSTTService

stt = SaidwireSTTService(
    settings=SaidwireSTTService.Settings(early_words=True),
)


@stt.event_handler("on_early_word")
async def on_early_word(stt, word):
    # {"word": "cancel", "punctuated_word": "cancel.", "start": 6.84, "end": 7.29, ...}
    ...

Notes

Audio format: The service sends 16-bit mono PCM at the pipeline’s input rate.
Idle connections: While no audio flows (for example, while the user is muted), the service sends KeepAlive messages instead of silent audio.
Errors: A bad or revoked key, or used-up credit, is reported once as a permanent error, with no retry loop.
Language: English only.

Compatibility

Tested with Pipecat 1.15.0 and Python 3.12. See the source repository for the latest tested version and changelog.

Get Started

Standards

Pipecat Server

Client SDKs

Pipecat Flows

Pipecat Cloud

CLI

Pipecat Context Hub

Saidwire Speech-to-Text

Overview

Source Repository

Example

Saidwire API Docs

PyPI Package

Installation

Prerequisites

Saidwire Account Setup

Required Environment Variables

Configuration

Settings

Usage

Basic Setup

With Early Words

Notes

Compatibility

​Overview

Source Repository

Example

Saidwire API Docs

PyPI Package

​Installation

​Prerequisites

​Saidwire Account Setup

​Required Environment Variables

​Configuration

​Settings

​Usage

​Basic Setup

​With Early Words

​Notes

​Compatibility

Overview

Installation

Prerequisites

Saidwire Account Setup

Required Environment Variables

Configuration

Settings

Usage

Basic Setup

With Early Words

Notes

Compatibility