# No code snippets available yet for this library.
# To use this model, check the repository files and the library's documentation.
# Want to help? PRs adding snippets are welcome at:
# https://github.com/huggingface/huggingface.jsAnima-2.9B
Unmodified mirror of Gazingstars123/Anima-2.9B (preview-v1) for JoyFusion's model downloads. Every file here is byte-identical to the upstream repository.
Anima-2.9B is a community fine-tune and layer expansion of circlestone-labs/Anima: the transformer was expanded from 28 to 40 blocks (about 2.9B parameters) and trained on additional anime and illustration data. It uses the same Qwen3-0.6B text encoder and Qwen-Image VAE as Anima 2B.
| File | Variant | Size | SHA-256 |
|---|---|---|---|
Anima-2.9B-preview-v1.safetensors |
BF16 | 5,843,204,206 | 0b3020d1b906155f7eb30667622723e87160632c8c7a5f1c93bdce685f2a346d |
Anima-2.9B-preview-v1_int8_convrot.safetensors |
INT8 (convrot) | 3,082,486,758 | cfb6ee2dde085de2b1c52b3fbe098dbc8c9ebd24f290c59755ab2c5f9291d396 |
config.json and expand_manifest.json are copied from upstream and describe the layer expansion.
Upstream recommends 28–50 steps, CFG 3.5–5 and resolutions such as 832×1216 or 1152×1536. There is no distilled (Turbo) variant yet.
License
Anima is licensed by CircleStone Labs LLC under the CircleStone Labs Non-Commercial License (see LICENSE.md), and
Anima-2.9B falls under its derivative-model category. Use and distribution are permitted for non-commercial purposes
only. See NOTICE for the required attribution notice.
- Downloads last month
- 5
Model tree for JoyFusionAI/Anima-2.9B
Base model
nvidia/Cosmos-Predict2-2B-Text2Image