Voxtral-TTS-Q4 Setup Prompt

Take the following sources and create a prompt for a coding agent which will set up voxtral-tts-q4.gguf https://mistral.ai/news/voxtral-tts https://huggingface.co/TrevorJS/voxtral-tts-q4-gguf#:~:text=Usage-,Native%20CLI,%2D%2D%20speak%20%2D%2Dlist%2Dvoices https://www.google.com/search?q=voxtral-tts-q4.gguf+set+up+instructions&oq=voxtral-tts-q4.gguf+set+up+instructions&gs_lcrp=EgZjaHJvbWUyCQgAEEUYORigATIHCAEQIRigAdIBCDQ4MjNqMGo3qAIAsAIA&sourceid=chrome&ie=UTF-8

Here’s a **ready-to-use coding agent prompt** that will set up **`voxtral-tts-q4.gguf`** (a quantized open-source text-to-speech model) on a development system, using the official Hugging Face model and native CLI instructions from the model card. This prompt is designed for a coding agent (e.g., an autonomous dev assistant, AI engineer bot, or script generator) so it *downloads, installs, and configures the model for inference*. --- ## 🧠 **Coding Agent Prompt — Setup Voxtral-TTS-Q4** **Context:** You are an autonomous coding agent. Your task is to set up the **Voxtral TTS Q4 GGUF model** locally for native inference and testing. The model is a quantized 4-bit TTS model in GGUF format with ~2.67 GB weights, suitable for text-to-speech synthesis on local machines. citeturn1view0 **Goal:** Write a script or automation that: 1. Downloads the `voxtral-tts-q4.gguf` model from Hugging Face. 2. Installs all required dependencies (Rust/WASM toolchains and any TTS runner like `voxtral-mini-realtime-rs` or `burn`). 3. Verifies the environment and model integrity. 4. Provides example commands to run the TTS CLI and list available voices. 5. Documents how to test synthesis and debug issues. --- ## 📌 **Prompt for the Coding Agent** ``` You are a coding agent tasked with fully automating the setup of the Voxtral TTS Q4 GGUF model for local native inference. Use the TrevorJS/voxtral-tts-q4-gguf model from Hugging Face. 1. **Environment Preparation** - Detect the host OS (Linux / macOS / Windows) and install or verify: * Rust toolchain (rustup, cargo) * wgpu dependencies for GPU support (Vulkan/Metal/DX12) * `huggingface_hub` CLI or Python package * Any build tools needed (e.g., CMake, build-essential) 2. **Model Download** - Use Hugging Face CLI or `huggingface_hub` to download `voxtral-tts-q4.gguf` into a local models directory: ``` hf download TrevorJS/voxtral-tts-q4-gguf voxtral-tts-q4.gguf --local-dir models ``` - Verify the file exists and is ~2.67 GB. 3. **Inference Runner** - Choose an inference runner (Rust-based `voxtral-mini-realtime-rs` or CLI with wgpu). - Clone and build the runner: ``` git clone https://github.com/TrevorJS/voxtral-mini-realtime-rs cd voxtral-mini-realtime-rs cargo build --release ``` - Link the runner to the downloaded GGUF model. 4. **Testing** - Provide commands to: * List available voices: ``` ./target/release/voxtral speak --list-voices --gguf models/voxtral-tts-q4.gguf ``` * Synthesize text: ``` ./target/release/voxtral speak --text "Hello world" --voice casual_female --gguf models/voxtral-tts-q4.gguf ``` * Use real-time settings (e.g., Euler steps). 5. **Documentation** - Output a README that explains how to: * Reproduce the setup * Switch voices * Debug common errors (e.g., missing wgpu drivers) * Integrate into Python/Node workflows 6. **Verification** - Confirm that TTS audio files are generated successfully (e.g., save output to WAV/MP3). - Print logs of success/failure. Produce a single script (Bash/Python/Makefile) and accompanying README text that accomplishes all of the above. ``` --- ## 🛠️ **Key Setup Instructions Embedded** - Model weights & quantization: **`voxtral-tts-q4.gguf`** is the quantized model file in GGUF format (~2.67 GB). citeturn1view0 - Native CLI usage example: ``` uv run --with huggingface_hub hf download TrevorJS/voxtral-tts-q4-gguf voxtral-tts-q4.gguf --local-dir models cargo run --release --features "wgpu,cli,hub" --bin voxtral -- speak --text "Hello world" --voice casual_female --gguf models/voxtral-tts-q4.gguf ``` *and* list voices: ``` cargo run --release --features "wgpu,cli,hub" --bin voxtral -- speak --list-voices ``` citeturn1view0 --- ## 📌 **What This Prompt Enables** ✅ Fully automated local setup of a state-of-the-art open-source TTS model ✅ Voice synthesis across 9 languages (English, French, Spanish, German, Italian, Portuguese, Dutch, Hindi, Arabic) ✅ Integration ready for CLI, app embedding, or production pipelines --- If you want, I can also generate a **ready-to-run shell script or Makefile** that implements this setup end-to-end.