Request for Official ONNX Export to Enable AMD GPU Acceleration on Windows
Hi IndexTeam,
Thank you very much for your impressive work on IndexTTS-2 and it is really an amazing TTS model i have seen so far!
I’d like to kindly request an official ONNX export of the model to support users on Windows 10/11 with AMD GPUs (such as the Radeon RX 580).
Since AMD’s consumer GPUs do not support ROCm—and CUDA is NVIDIA-only—most deep learning frameworks fall back to CPU inference, significantly limiting performance. However, ONNX Runtime with the DirectML execution provider enables GPU-accelerated inference on DirectX 12–compatible AMD hardware(including mine RX 580) via standard Windows drivers.
Providing an ONNX version would unlock accessible, hardware-accelerated TTS for a broader community of Windows + AMD users without requiring specialized compute stacks. Please kindly request an official ONNX export of the model.