More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Qwen3-TTS Local Voice

    1

    Local character voice generation and batch audio production using Qwen3-TTS and MLX on Apple Silicon.

    $3.99

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Qwen3-TTS Local Voice

    Qwen3-TTS Local Voice

    Example session with this skill installed

    Generate a voice design and audition for a character named Eara, described as a calm, middle-aged elven woman with a melodic, slow pace in English.

    • Read your context and instructions
    • Compiled the qwen3-tts local voice

    I have created the voice design for 'Eara' and queued the audition generation. The design file has been updated in examples/voice-designs.example.json and the audio file is ready for review.

    Voice Design and Audition Results

    Role IDDescriptionAudition LineStatus
    eara_01Middle-aged female, calm, melodic timbre, slow deliberate pace, elven, English."The forest remembers what the stone has long forgotten, traveler."Audition Generated

    Design Specification Summary:
    The instruction set was cleaned to remove abbreviations and production notes to prevent prompt leakage. The model Qwen3-TTS-12Hz-1.7B-VoiceDesign-8bit is being used to ensure local execution.

    Audio Asset Path:
    PROJECT_DIR/assets/audio/voice/auditions/eara_01.mp3

    Next steps

    1. Run python scripts/transcribe_audit.py --audio assets/audio/voice/auditions/eara_01.mp3 to verify the clip contains no leaked instructions.
    2. Listen to the MP3 to confirm the melodic timbre matches the character profile.
    3. Add the role to your manifest file to begin batch dialogue generation.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Cloud-based text-to-speech services require constant connectivity and expensive API keys. This skill enables local generation of high-quality character voice assets on Apple Silicon using Qwen3-TTS.

    What it does

    • Voice design creates unique character voices from natural language descriptions of age, timbre, and emotion.
    • Batch generation produces fixed WAV or MP3 assets for game dialogue and interactive stories.
    • Audio auditing uses local Whisper models to verify transcripts and catch prompt leakage before production.
    • Resource management manages 8-bit MLX model downloads and memory footprints for 16GB M-series Macs.
    • Manifest orchestration coordinates role assignments and scene-based line generation via structured JSON schemas.

    How it works

    1. Initialize the isolated MLX environment and download the 8-bit VoiceDesign model to local storage.
    2. Define character roles in a design file and dialogue lines in a project manifest.
    3. Run the audition command to generate and review sample clips for each character role.
    4. Execute batch generation to produce the full set of fixed audio assets for the application.

    Frameworks & tools

    This skill uses Python 3.13, MLX, and ffmpeg. It is optimized for Apple Silicon (M1/M2/M3/M4) running macOS.

    Why this beats prompting it yourself

    Manual prompting often leads to prompt leakage where the model speaks the instructions or markup tags. This skill implements a strict workflow that strips metadata and calls the VoiceDesign model directly, preventing mixed-language artifacts and drift.

    Use cases

    • Generate hundreds of unique dialogue lines for an offline visual novel.
    • Create character auditions for role-playing games without paying per-token TTS fees.
    • Produce consistent voice assets for audio dramas using local hardware.
    • Rapidly prototype character voices using natural language descriptions instead of reference clips.

    Known limitations

    Requires an M-series Mac with at least 16GB RAM. VoiceDesign models may drift between calls; use the base voice-clone model for long-term stability across large asset sets.

    Requires macOS on Apple Silicon. Needs roughly 6GB free memory for the 8-bit model. ffmpeg must be installed for MP3 conversion.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 2 days ago

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Listed2 days ago

    What's inside

    Frequently Asked Questions

    This skill is built for Apple Silicon (M1/M2/M3/M4) devices running macOS. It requires a minimum of 16GB of Unified Memory to effectively manage the 8-bit MLX model and local inference processes.

    Unlike standard prompting which often results in the model reading aloud its own instructions, this skill uses a structured manifest system and post-generation auditing to ensure only the intended dialogue is captured in the final audio assets.

    The package includes the Python-based orchestration layer, MLX environment configurations, and pre-defined JSON schemas for role assignment. It does not include the Qwen3-TTS model weights, but it provides a secure automated script to download the optimized 8-bit versions to your local machine.

    While the skill excels at 'Voice Design' from descriptions, it also supports voice-cloning via reference clips. We recommend using the cloning mode for long-term projects to ensure character consistency across different recording sessions.

    This skill is a standalone local pipeline that outputs standard WAV or MP3 files. These assets can be imported into any game engine, interactive storytelling agent, or audio production suite that accepts standard audio formats.

    New in Game Development