Interested in integrating Airy as a TTS provider

[fbdp1202]

Hi! We'd like to add Airy as a TTS provider to the Coval TTS benchmark.

Airy provides a streaming TTS API through its Cloud API

We've implemented and tested an integration against the existing Coval provider interface.

Integration details

Field	Value
Provider	airy
Model	airy-tts-v1
Protocol	HTTP streaming
Endpoint	https://api.airy.so/v1/audio/speech/stream
Auth	AIRY_API_KEY
Audio	PCM S16LE, 24 kHz mono
Output	WAV, finalized from the streamed PCM response

The integration streams audio over HTTP and records time-to-first-audio while preserving the existing TTS benchmark output and diagnostics behavior.

It also validates the returned audio metadata and handles failed, interrupted, or incompatible streams.

Implementation

I've opened PR #608 with the provider integration.

It includes:

• Airy streaming TTS provider implementation
• Provider registry/configuration updates
• AIRY_API_KEY environment variable support
• Unit tests covering requests, audio output, timing, authentication, and error handling

Validation

The implementation has been tested with:

• 17 Airy provider unit tests passing
• Ruff lint/format checks passing
• mypy --strict src tests passing
• Full pytest suite passing
• Live API smoke test returning HTTP 200 over HTTP/2 with valid 24 kHz mono 16-bit WAV output

Happy to provide any additional API access or test credentials needed for benchmark validation.

Opening this issue to track the provider addition and discussion alongside PR #608.

Thanks!