Function run_vibevoice_1_5b

cake-cli/src/main.rs:440–561

(
    ctx: Context,
    model_path: &std::path::Path,
    config_path: &std::path::Path,
)

#[cfg(feature = "master")]
#[cfg(feature = "vibevoice")]
async fn run_vibevoice_1_5b(
    ctx: Context,
    model_path: &std::path::Path,
    config_path: &std::path::Path,
) -> Result<()> {
    use cake_core::models::vibevoice;
    use cake_core::models::vibevoice::config_1_5b::*;

    println!("[VibeVoice-1.5B] Loading from {}", model_path.display());

    // Collect weight shard paths
    let mut weight_paths: Vec<std::path::PathBuf> = Vec::new();
    for entry in std::fs::read_dir(model_path)? {
        let entry = entry?;
        let name = entry.file_name().to_string_lossy().to_string();
        if name.ends_with(".safetensors") && name.starts_with("model") {
            weight_paths.push(entry.path());
        }
    }
    weight_paths.sort();
    println!(
        "[VibeVoice-1.5B] Found {} weight shards",
        weight_paths.len()
    );

    let mut model = vibevoice::VibeVoice1_5B::load(
        config_path,
        &weight_paths,
        &ctx.device,
        Some(ctx.args.tts_diffusion_steps),
        &ctx.topology,
        ctx.args.cluster_key.as_deref(),
    )
    .await?;

    let prompt = &ctx.args.prompt;
    println!("[VibeVoice-1.5B] Generating speech for: \"{}\"", prompt);

    // Load tokenizer (Qwen2.5-1.5B)
    let tokenizer = {
        let local = model_path.join("tokenizer.json");
        if local.exists() {
            tokenizers::Tokenizer::from_file(&local)
                .map_err(|e| anyhow::anyhow!("tokenizer: {e}"))?
        } else {
            println!("[VibeVoice-1.5B] Downloading Qwen2.5-1.5B tokenizer...");
            let qwen_path = utils::hf::ensure_model_downloaded("Qwen/Qwen2.5-1.5B")?;
            tokenizers::Tokenizer::from_file(qwen_path.join("tokenizer.json"))
                .map_err(|e| anyhow::anyhow!("tokenizer: {e}"))?
        }
    };

    // Load voice reference audio
    let voice_path = ctx
        .args
        .voice_prompt
        .as_ref()
        .ok_or_else(|| anyhow::anyhow!("--voice-prompt required (path to .wav voice reference)"))?;