Trendyol TTS MLX Weights

MLX safetensors artifacts for the Apple Silicon / MLX port of Trendyol/Trendyol-TTS.

Code repository: https://github.com/emre-koc/trendyol-tts-mlx

Files

  • voxcpm2_layers_mlx.safetensors and voxcpm2_layers_manifest.json: exported VoxCPM2 local encoder, language-model, residual-LM, DiT/CFM, FSQ/projection, and stop-predictor weights/metadata for the strict MLX runtime.
  • audiovae_decoder_mlx.safetensors and audiovae_decoder_manifest.json: exported AudioVAE/vocoder decoder weights/metadata for MLX.

Usage

git clone https://github.com/emre-koc/trendyol-tts-mlx.git
cd trendyol-tts-mlx
/opt/homebrew/bin/python3.11 -m venv .venv
source .venv/bin/activate
pip install -e .

hf download emrekoc/trendyol-tts-mlx --local-dir models/Trendyol-TTS-mlx

trendyol-tts \
  --text "Merhaba, Trendyol TTS modelinden Türkçe bir ses örneği dinliyorsunuz." \
  --out out.wav \
  --device mps \
  --backend mlx \
  --cfg 2.0 \
  --steps 16 \
  --max-len 4096

The runtime is currently strict text-only MLX. Prompt/reference audio, streaming, denoiser/enhancer, and voice-cloning paths are intentionally unsupported until explicitly ported.

Downloads last month

-

Downloads are not tracked for this model. How to track
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support