pipertts - v1.3.0
    Preparing search index...

    Interface PiperInferenceOptions

    Full set of inference parameters accepted by PiperTTS.

    interface PiperInferenceOptions {
        configPath?: string;
        jsonInput?: boolean;
        lengthScale?: number;
        logLevel?: "debug" | "info" | "warn" | "error";
        modelPath?: string;
        noiseScale?: number;
        noiseWScale?: number;
        numThreads?: number;
        outputFile?: string;
        outputFormat?: PiperOutputFormat;
        sentenceSilence?: number;
        speakerId?: number;
        timeoutMs?: number;
        useCuda?: boolean;
    }
    Index
    configPath? jsonInput? lengthScale? logLevel? modelPath? noiseScale? noiseWScale? numThreads? outputFile? outputFormat? sentenceSilence? speakerId? timeoutMs? useCuda?
    configPath?: string

    Path to the model configuration JSON file (.onnx.json).

    jsonInput?: boolean

    Enable Piper JSON phoneme input mode.

    lengthScale?: number

    Speech rate multiplier (higher = slower). Range: 0.1 - 10.0.

    logLevel?: "debug" | "info" | "warn" | "error"

    Piper log verbosity. The Piper CLI only exposes a --debug flag, so only "debug" changes the spawned arguments. Other levels are accepted for forward-compatibility and recorded in the result options.

    modelPath?: string

    Path to the .onnx voice model file.

    noiseScale?: number

    Controls audio variability / expressiveness. Range: 0.0 - 2.0.

    noiseWScale?: number

    Controls duration/timing variability. Range: 0.0 - 2.0.

    numThreads?: number

    Number of ONNX inference threads. Must be an integer >= 1.

    outputFile?: string

    Output file path written by Piper.

    outputFormat?: PiperOutputFormat

    Output audio format.

    • wav: standard WAV file (default).
    • raw: raw PCM captured from stdout (no file written by Piper).
    • mp3 / ogg / opus: transcoded from WAV (ffmpeg when available, pure-JS fallbacks otherwise).
    sentenceSilence?: number

    Silence appended after each sentence, in seconds. Range: 0.0 - 10.0.

    speakerId?: number

    Speaker ID for multi-speaker models. Must be an integer >= 0.

    timeoutMs?: number

    Per-call synthesis timeout in milliseconds (positive integer).

    useCuda?: boolean

    Use CUDA GPU acceleration if available.

    Settings
    Member Visibility

    On This Page
    Properties