celeste-speech-generation 0.2.3

Type-safe speech generation for Celeste AI. Unified interface for Google and more

Celeste Speech Generation

Speech Generation capability for Celeste AI

Quick Start • Parameters • Request Provider

---

🚀 Quick Start

from celeste import create_client, Capability, Provider

# Create client
client = create_client(
    capability=Capability.SPEECH_GENERATION,
    provider=Provider.GOOGLE,
)

# Generate audio from text
response = await client.generate(prompt="Hello, this is a test.")
print(response.content)  # AudioArtifact with binary audio data

Install:

uv add "celeste-ai[speech-generation]"

---

🔄 Multi-Provider Support

# Switch providers with zero code changes
from celeste import create_client, Capability, Provider

# After implementing multiple providers, switch by changing one parameter:
client_a = create_client(Capability.SPEECH_GENERATION, Provider.OPENAI)
client_b = create_client(Capability.SPEECH_GENERATION, Provider.GOOGLE)
client_c = create_client(Capability.SPEECH_GENERATION, Provider.ELEVENLABS)

# Same method, same parameters, different results:
response_a = await client_a.generate(prompt="Your text", voice="alloy")
response_b = await client_b.generate(prompt="Your text", voice="Kore")
response_c = await client_c.generate(prompt="Your text", voice="default")

One API. Multiple providers. Zero vendor lock-in.

---

Supported Providers

Provider	Status	Models
Google	✅ Implemented	gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts

Missing a provider? Request it – ⚡ we ship fast.

Use /implement-provider to add provider support.

---

Parameters

💡 Parameter Mapping: Celeste automatically translates unified parameter names to each provider's native format. Write voice once, we handle provider-specific voice IDs.

Parameter	Type
-	-

Parameters will be added as providers are implemented. Use /add-parameters to add parameter support.

Full parameter reference: API Docs

---

Streaming: ✅ Supported (3/4 providers confirmed)

Streaming support based on cross-provider research (OpenAI, ElevenLabs, Amazon Polly).

---

Examples

Basic Generation

# Generate audio from text
response = await client.generate(prompt="Hello, this is a test.")

# Access audio artifact
audio = response.content  # AudioArtifact
audio.mime_type  # "audio/mp3"
audio.data  # Binary audio bytes

Streaming Audio

# Stream audio as it's generated
stream = client.stream(prompt="This is a longer text to synthesize.")

async for chunk in stream:
    # Process audio chunks
    audio_bytes = chunk.content  # Raw bytes
    # Play or save chunks as they arrive

Switch Providers

# After multiple providers implemented
client = create_client(Capability.SPEECH_GENERATION, Provider.OPENAI)
response = await client.generate(prompt="...", voice="nova", speed=1.2)

---

🤝 Contributing

See CONTRIBUTING.md for guidelines.

Request a provider: GitHub Issues

---

📄 License

Apache 2.0 License – see LICENSE for details.

---

Get Started • Documentation • GitHub

Made with ❤️ by developers tired of framework lock-in