Crate autoagents_speech 

AutoAgents Speech

Speech (TTS/STT) provider abstractions for the AutoAgents framework.

This crate provides trait-based abstraction layers for speech providers, allowing different backends to be used interchangeably within the AutoAgents ecosystem.

Features

TTS (Text-to-Speech)

• Speech Generation: Generate audio from text
• Voice Management: Use predefined voices
• Streaming Support: Optional streaming for real-time audio generation
• Model Management: Support for multiple models and languages

STT (Speech-to-Text) - Coming Soon

Architecture

The crate follows a trait-based design with provider implementations in the providers module:

TTS Traits

• TTSProvider: Marker trait combining all TTS capabilities
• TTSSpeechProvider: Speech generation capabilities
• TTSModelsProvider: Model and language support

Providers

Enable providers using feature flags:

• pocket-tts: Pocket-TTS model support

Example

use autoagents_speech::{TTSProvider, SpeechRequest, VoiceIdentifier, AudioFormat};

async fn generate_speech(provider: &dyn TTSProvider, text: &str) {
    let request = SpeechRequest {
        text: text.to_string(),
        voice: VoiceIdentifier::new("alba"),
        format: AudioFormat::Wav,
        sample_rate: Some(24000),
    };

    let response = provider.generate_speech(request).await.unwrap();
    println!("Generated {} samples", response.audio.samples.len());
}

Re-exports

pub use error::TTSError;

pub use error::TTSResult;

pub use types::AudioChunk;

pub use types::AudioData;

pub use types::AudioFormat;

pub use types::ModelInfo;

pub use types::SharedAudioData;

pub use types::SpeechRequest;

pub use types::SpeechResponse;

pub use types::VoiceIdentifier;

Modules

error

providers

Speech provider implementations

types

Traits

TTSModelsProvider

Trait for TTS model management capabilities

TTSProvider

Marker Trait for TTS providers

TTSSpeechProvider

Trait for TTS speech generation capabilities