cicalooo's picture
Add example workflow and screenshot to model card
8120b0b verified
|
Raw History Blame Contribute Delete
1.58 kB
---
license: other
license_name: fish-audio-community-license
license_link: LICENSE
base_model: fishaudio/drama-3-int4
base_model_relation: quantized
pipeline_tag: text-to-speech
tags: [fish-audio, text-to-speech, comfyui, int4, safetensors]
---
# Drama 3 int4 single-file (ComfyUI)
Single-file repack of [fishaudio/drama-3-int4](https://huggingface.co/fishaudio/drama-3-int4) (int4 quant of [fishaudio/drama-3](https://huggingface.co/fishaudio/drama-3)) for a local ComfyUI node pack.
- `drama3_int4.safetensors`: all 12 shards merged; config, tokenizer, chat template and quant config stored in safetensors metadata. Tensors byte-identical to the source.
- `drama3_codec.safetensors`: Fish S2 codec (44.1 kHz) converted from `codec.pth` (6 non-persistent buffers dropped).
Place both in `ComfyUI/models/TTS/`.
## Example workflow
![FishDramatics example workflow](example_workflow.png)
Drag [`example_workflow.json`](example_workflow.json) into ComfyUI to load it. The node pack lives in [`ComfyUI-FishDramatics/`](ComfyUI-FishDramatics/); install just the nodes from inside `ComfyUI/custom_nodes` with:
hf download cicalooo/ComfyUI-FishDramatics --include "ComfyUI-FishDramatics/*" --local-dir .
## License
Derivative Work of Fish Audio Materials under the Fish Audio Community License v1.0 ([LICENSE](LICENSE)): research/non-commercial use free; commercial use only under USD 500k annual revenue; no cloning real people's voices without consent; do not use outputs to train other speech or foundation models.
Credits: Fish Audio (model), AK (int4 quantization).