Connect Kitta Audio TTS to Pipecat

Pipecat's OpenAITTSService accepts a custom base_url, allowing Kitta Audio to provide speech in a realtime voice pipeline.

Install and configure

uv add "pipecat-ai[openai]"

import os
from pipecat.services.openai import OpenAITTSService

tts = OpenAITTSService(
    api_key=os.environ["FISH_AUDIO_API_KEY"],
    base_url="https://fishaudio.org/v1",
    settings=OpenAITTSService.Settings(
        model="fishaudio-s21pro-flash",
        voice="00a1b221-6137-4b73-ad62-b0cbce134167",
        speed=1.0,
    ),
)

Add tts to the Pipecat pipeline and verify that a short sentence reaches the transport as audio frames before adding STT, an LLM, and interruption handling.

Audio considerations

Pipecat's OpenAI TTS pipeline processes 24 kHz PCM audio. Do not stop at an HTTP 200 check: listen to the result and verify sample rate, channel count, and playback speed.

Troubleshooting

• Deprecated parameter warning: use OpenAITTSService.Settings rather than legacy top-level model and voice arguments.
• Fast, slow, or noisy audio: align the transport sample rate with the TTS service.
• 401/402: verify the server-side environment variable and API credits.

Related documentation

• Pipecat OpenAI TTS
• API errors