Args:
        text: The text to synthesize into speech.
        language: Target language (ISO code or 'Auto'). 646 languages
            supported. Omit to use the voice profile's saved language;
            an explicit 'Auto' overrides it.
        profile_id: ID of a saved voice profile to clone. Omit to use this
            agent's bound voice (Settings → MCP), else the default voice.
        instruct: Style instruction (e.g. 'whisper', 'excited', 'narrator').
        speed: Speech speed multiplier (0.5–2.0, default 1.0).
        steps: Diffusion steps (8=fast/draft, 16=balanced, 32=quality).
        format: File and URL format: wav (default), ogg or opus. Both
            ogg and opus carry Opus in Ogg; requires files/both mode and ffmpeg.

    Returns:
        JSON with audio_id, generation_time_s, audio_duration_s and the
        audio shaped by OMNIVOICE_MCP_OUTPUT_MODE: base64 WAV data
        ('resources', the default), a URL plus an optional file ('files'),
        or both ('both'). Prefer 'files' for LLM agents.
     For long operations use voicestudio_start_job.

    Returns presets like Narrator, Casual, News Anchor, etc. with their
    instruct text. Use the instruct text with generate_speech.
    

    Pass exactly one of audio_base64 or audio_path.

    Args:
        audio_base64: Base64-encoded audio bytes (wav/mp3/webm/m4a).
        audio_path: Path to an audio file under OMNIVOICE_MCP_BASE_PATH
            (relative to it, or absolute inside it). The base path is the
            security boundary: with none configured, paths are refused.
            Prefer this lane for LLM agents - the audio never enters the
            agent's context.
        language: Optional language hint; omit for auto-detect.

    Returns:
        JSON with the recognized text, language, and duration.
     For long operations use voicestudio_start_job.

    The new voice is immediately available for use with generate_speech
    (pass the returned profile_id as the profile_id argument). Pass
    exactly one of ref_audio_base64 or ref_audio_path.

    Args:
        name: A human-friendly name for the cloned voice.
        ref_audio_base64: Base64-encoded audio (WAV, MP3, FLAC, etc.) of
            the reference voice — 5-30 seconds of clean single-speaker
            speech.
        ref_text: Optional transcript of the reference audio (improves
            quality for some engines).
        instruct: Optional style instruction (e.g. 'whisper', 'excited').
        language: Language of the reference audio (ISO code or 'Auto').
        ref_audio_path: Path to the reference audio under
            OMNIVOICE_MCP_BASE_PATH (relative to it, or absolute inside
            it); refused when no base path is configured. Prefer this
            lane for LLM agents - the clip never enters the context.

    Returns:
        JSON with the new profile's id, name, and kind.
     For long operations use voicestudio_start_job.

    Nothing is saved. Use this before design_voice to see what the
    description will produce. The design space is small and fixed; only
    these tokens (and close synonyms) are understood:
      Gender: male, female
      Age: child, teenager, young adult, middle-aged, elderly
      Pitch: very low / low / moderate / high / very high pitch
      Style: whisper
      EnglishAccent: american, british, australian, canadian, indian,
        japanese, korean, chinese, russian, portuguese accent
      ChineseDialect (Chinese speech; overrides an English accent):
        sichuan, dongbei / northeastern chinese, henan, shaanxi, gansu,
        guilin, guizhou, jinan, ningxia, qingdao, shijiazhuang, yunnan
        (e.g. "sichuan dialect")
    Timbre words ("gravelly", "raspy") and other accents are ignored and
    reported in `unmatched`. For a voice outside this space, use
    clone_voice with reference audio instead.

    Args:
        description: Free-text description, e.g. "an elderly man with a
            deep voice and a british accent".

    Returns:
        JSON with attrs (category → token or "Auto"), instruct, matched
        and unmatched.
    

    The description is mapped onto the same attributes describe_voice
    previews (see its docstring for the vocabulary). The backend tries to
    render a fixed-seed identity sample at save time; if the voice engine
    isn't ready, the profile is still saved and the same sample is
    rendered on first use, so the voice stays stable across
    generate_speech calls either way. Pass the returned profile_id to
    generate_speech. Refuses a description that matches no attribute.

    Args:
        name: A human-friendly name for the new voice.
        description: Free-text description of the voice.
        language: The voice's saved language (ISO code or 'Auto'); used
            for its sample and by generate_speech calls that omit one.

    Returns:
        JSON with the new profile's id, name, kind, the attrs used, and
        any unmatched description fragments.
     For long operations use voicestudio_start_job.