Function run_master_luxtts

cake-cli/src/main.rs:578–619

(mut ctx: Context)

#[cfg(feature = "master")]
#[cfg(feature = "luxtts")]
async fn run_master_luxtts(mut ctx: Context) -> Result<()> {
    use cake_core::models::luxtts;
    use cake_core::models::Generator;

    println!("[LuxTTS] Loading model from {}...", ctx.data_path.display());

    let mut model = luxtts::LuxTTS::load(&mut ctx)
        .await?
        .ok_or_else(|| anyhow::anyhow!("failed to load LuxTTS model"))?;

    let prompt = ctx.args.prompt.clone();
    println!("[LuxTTS] Generating speech for: \"{}\"", prompt);

    // Load reference audio if provided
    let reference_audio = if let Some(ref ref_path) = ctx.args.tts_reference_audio {
        println!("[LuxTTS] Loading reference audio: {}", ref_path);
        Some(load_wav_mono_24k(std::path::Path::new(ref_path))?)
    } else {
        None
    };

    let samples: Vec<f32> = model
        .generate_speech(
            &prompt,
            reference_audio.as_deref(),
            ctx.args.tts_t_shift,
            ctx.args.tts_cfg_scale,
            ctx.args.tts_diffusion_steps.min(10), // LuxTTS uses 4 steps by default
            ctx.args.tts_speed,
        )
        .await?;

    let output_path = std::path::Path::new(&ctx.args.audio_output);
    luxtts::save_wav(&samples, output_path, 48000)?;
    println!(
        "[LuxTTS] Audio saved to {} ({:.1}s, {} samples @ 48kHz)",
        output_path.display(),
        samples.len() as f64 / 48000.0,
        samples.len()
    );
    Ok(())
}

#[cfg(feature = "master")]
async fn run_master_image(ctx: Context) -> Result<()> {