pipecat-plugins-pinch

Real-time voice translation for Pipecat powered by Pinch.

Installation

pip install pipecat-plugins-pinch

Requires Python ≥ 3.10.

Prerequisites

You need a Pinch API key. Get one at the developers portal.

Set it in your environment:

export PINCH_API_KEY=pk_your_key_here

How it fits into a Pipecat pipeline

PinchTranslatorService is a drop-in FrameProcessor. It sits between your transport's input and output — receiving InputAudioRawFrame from the user and emitting OutputAudioRawFrame (translated speech) and TranscriptionFrame (transcripts) downstream.

transport.input()
      │
      ▼
PinchTranslatorService        ← translate en-US → es-ES
      │
      ├─► OutputAudioRawFrame  → transport.output()  (translated audio)
      └─► TranscriptionFrame   → your handler        (transcripts)

Usage

from pipecat.pipeline.pipeline import Pipeline
from pipecat.pipeline.runner import PipelineRunner
from pipecat.pipeline.task import PipelineParams, PipelineTask
from pipecat.plugins.pinch import PinchTranslatorService, TranslatorOptions

async def main():
    # Replace with your preferred Pipecat transport (Daily, LiveKit, WebSocket, etc.)
    # transport = DailyTransport(...)

    translator = PinchTranslatorService(
        options=TranslatorOptions(
            source_language="en-US",
            target_language="es-ES",
            voice_type="clone",   # preserves the speaker's voice
        ),
        # api_key="pk_..."  ← or set PINCH_API_KEY env var
    )

    pipeline = Pipeline([
        transport.input(),
        translator,
        transport.output(),
    ])

    task = PipelineTask(pipeline, params=PipelineParams(allow_interruptions=True))
    await PipelineRunner().run(task)

The plugin handles everything internally — calling the Pinch API, managing the translation session, resampling audio, and routing frames — so you don't need to configure anything beyond TranslatorOptions.

Configuration

TranslatorOptions

PinchTranslatorService

Supported languages

Full list of language codes: supported languages

Transcript frames

The plugin emits standard Pipecat transcript frames for both the original and translated speech. You can consume these in any downstream FrameProcessor.

from pipecat.frames.frames import TranscriptionFrame, InterimTranscriptionFrame
from pipecat.processors.frame_processor import FrameProcessor, FrameDirection

class MyTranscriptHandler(FrameProcessor):
    async def process_frame(self, frame, direction):
        await super().process_frame(frame, direction)

        if isinstance(frame, TranscriptionFrame) and frame.user_id == "":
            print(f"Translated: {frame.text}")

        await self.push_frame(frame, direction)

Links

• Pinch
• Pipecat documentation
• Pipecat GitHub

License

Apache 2.0 — see LICENSE.