When TTS-Audio-Suite updates, python packages are downgraded

[jason-mightynice]

Downgrades

transformers-4.57.3
huggingface-hub-0.36.2

onnxruntime-1.24.4 gets reinstalled when onnxruntime-gpu is already installed (gpu accelerated for DW POSE).

[*] Installing transformers>=4.51.3,<=4.57.3.

I use the latest which i need to be able to use ComfyUI-OmniVoice-TTS (ComfyUI-LongCat-AudioDIT-TTS also works if you are looking to support).

I re-update everything and it's all good until there is a TTS-Audio-Suite update.

[diodiogod]

I fixed the ONNX Runtime installer behavior in the latest update.

What changed:

• TTS Audio Suite now detects an existing onnxruntime-gpu install and keeps it instead of reinstalling CPU onnxruntime
• The Python 3.13 OpenSeeFace path no longer force-reinstalls CPU onnxruntime over a working GPU setup

Please update and test again in your ComfyUI environment.

For the transformers part: I need the exact version you require.

• If you mean transformers 5.x, that is intentionally unsupported right now because it breaks compatibility in TTS Audio Suite
• If you need something newer than our current upper bound but still within 4.x, tell me the exact version and which node requires it so I can evaluate whether that range can be expanded safely

[jason-mightynice]

Thank you for looking at it. Yes, I have py3.13 so the onnxruntime force reinstall fix is great.

yes, the transformers is the latest release 5.5.0.

your audio suite has so many models so i am sure compatibility is tricky. (i had to update initially for omnivoice). I would prefer if the node didn't downgrade my transformers in my comfyui environment . I usually use the latest SOTA TTS node, and when it comes to TTS-Audio_Suite, I move to that.

[diodiogod]

I got exhausted trying to patch transformers 5.0.0 at the time of it's release. So I gave up #238
But I should try again in the future...

In the meantime, I recommend using a portable installation just for TTS Audio Suite, or for the nodes that requires transformers 5.5