Irodori-TTS-v4.1-q8_0.gguf a try! Converted from Aratako/Irodori-TTS-v4.1-Small and successfully audio-tested using audio.cpp. 🔥

[datkun92-code]

@0xShug0 Give Irodori-TTS-v4.1-q8_0.gguf a try! Converted from Aratako/Irodori-TTS-v4.1-Small and successfully audio-tested using audio.cpp. 🔥

📦 Download mirror: https://drive.google.com/file/d/1gWHP2g2a_F-CLGs3Vu5HIBDm0vpRrn2l/view

💡 Key highlight: VRAM stays under 2GB on my dac-vae patch decoding seamlessly on Vulkan, CUDA, Metal, and CPU backends.

Feel free to re-upload it to Hugging Face for the community! 🚀

[0xShug0]

Thanks @datkun92-code! BTW I will revisit your patch later.

[0xShug0]

I’ve updated both the f16 and q8 models. They still use the old names, but they’re actually v4.1.

@datkun92-code I couldn't find the comment that has your patch...

[datkun92-code]

@0xShug0 Please share the

text_encoder_config.json and model_config.json

files from your Irodori-TTS-v4.1-Small project.

[0xShug0]

They are generated by the official script in the Irodori-TTS repo convert_checkpoint_to_safetensors.py.

text_encoder_config.json

model_config.json