add hume tts integration

[ingridxx]

TTS request is valid but missing output file when running test_hume_tts.py

[markbackman]

You can leave this file blank. The pattern is to import from:

from pipecat.services.hume.tts import HumeTTSService

in application code.

[markbackman]

Nit: Maybe define this outside of the HumeTTSService class for ease of use?

[markbackman]

Do you support wav?

[markbackman]

You don't want to se the sample rate here. Instead, you want it to be set to None.

I wrote the following guidance recently on how this works:

### Sample Rate Handling

Sample rates are set via PipelineParams and passed to each frame processor at initialization. The pattern is to _not_ set the sample rate value in the constructor of a given service. Instead, use the `start()` method to initialize sample rates from the frame:

```python
async def start(self, frame: StartFrame):
    """Start the service."""
    await super().start(frame)
    self._settings["output_format"]["sample_rate"] = self.sample_rate
    await self._connect()

Note that self.sample_rate is a @property set in the TTSService base class, which provides access to the private sample rate value obtained from the StartFrame.

Check out other TTS services to see how the pattern is applied.

[markbackman]

For an HTTP based TTS service, you want the following:

super().__init__(sample_rate=sample_rate, **kwargs)

• aggregate_sentences: defaults to True, so no need to set
• push_text_frames: the TTS service should push text frames downstream. This is required for context aggregation. Default is True.
• push_stop_frames: You want the HumeTTSService to yield the frame instead. False is the default and desired value.

[markbackman]

In other services, the base_url is an arg. It could make sense here, too, for flexibility.

[markbackman]

Remove since this is unused.

Suggested change

self._cumulative_time = 0

[markbackman]

Remove since this is unused.

Suggested change

self._started = False

[markbackman]

From the TTSService class, there's a chunk_size @property.

    @property
    def chunk_size(self) -> int:
        """Get the recommended chunk size for audio streaming.

        This property indicates how much audio we download (from TTS services
        that require chunking) before we start pushing the first audio
        frame. This will make sure we download the rest of the audio while audio
        is being played without causing audio glitches (specially at the
        beginning). Of course, this will also depend on how fast the TTS service
        generates bytes.

        Returns:
            The recommended chunk size in bytes.
        """
        CHUNK_SECONDS = 0.5
        return int(self.sample_rate * CHUNK_SECONDS * 2)  # 2 bytes/sample

We've found this work well to avoid audio glitches in playback. It's helpful to use the property so we can uniformly adjust all HTTP based services.

[markbackman]

Instead of a test, can you write a foundational example. Follow the 07-interruptible pattern.

[markbackman]

Nice start! Sorry for the delayed feedback. We've been busy.

Four other things to do:

1. Update the README with your service.
2. Add env var to the env.example.
3. Add docs: https://github.com/pipecat-ai/docs
4. Make sure to lint your code. You can install the pre-commit hook using uv run pre-commit install from the base of the repo.

[markbackman]

Oops. Closing in favor of:
#2518