MLX IndexTTS 2.5

Pure MLX FP16 inference checkpoint for AI2Apps on Apple Silicon. The checkpoint is reproducibly converted from the pinned upstream IndexTTS 2.5, W2V-BERT, CAMPPlus, and BigVGAN revisions recorded in conversion-manifest.json.

Supported in AI2Apps 0.1.0: Chinese and English synthesis, one reference voice, the upstream eight-axis numeric emotion representation, and native duration control. The optional Qwen text-to-emotion model and Torch training state are not included.

Use is governed by LICENSE-IndexTTS-2.5.txt. The vendored MLX implementation is attributed under LICENSE-WIndexTTS.txt. This conversion is not endorsed by the upstream authors.

Downloads last month
-
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support