I created custom nodes for Fish Audio Drama 3 - Works with as little as 8GB VRAM - Quantized weights on HF
retained text
[**Fish Audio Drama 3**](https://huggingface.co/fishaudio/drama-3), converted for the native ComfyUI nodes of [**ComfyUI-Drama3**](https://github.com/PedroMarinhoDev/ComfyUI-Drama3): a 35B text-to-speech model for voice cloning, dialogue and inline cues such as `[sighs]` or `[whispers]`, running **on cards from 8 GB**. 90% of the model is routed experts and each audio frame uses only 8 of the 256 in a layer, so the nodes keep the experts in system RAM and a cache of the ones in use on the GPU. Each file holds the whole model in one file; the codec is separate. Weights: [https://huggingface.co/PedroMarinhoDev/Drama-3-ComfyUI](https://huggingface.co/PedroMarinhoDev/Drama-3-ComfyUI) Nodes: [https://github.com/PedroMarinhoDev/ComfyUI-Drama3](https://github.com/PedroMarinhoDev/ComfyUI-Drama3) Demo page: [https://pedromarinhodev.github.io/ComfyUI-Drama3/](https://pedromarinhodev.github.io/ComfyUI-Drama3/) Workflows: [https://github.com/PedroMarinhoDev/ComfyUI-Drama3/tree/main/example\_workflows](https://github.com/PedroMarinhoDev/ComfyUI-Drama3/tree/main/example_workflows) Let me know if yo have any issues. # [](https://huggingface.co/PedroMarinhoDev/Drama-3-ComfyUI#samples)
document fetched 2026-10-15 00:34 · canonical https://reddit.com/r/StableDiffusion/comments/1x5ynlf/i_created_custom_nodes_for_fish_audio_drama_3