Anima Extra β Karume
This is an unofficial conversion of WAI-ANIMA v1.0 (base 1.0) and copycat-anima 20260610 for Karume. It is not affiliated with or endorsed by WAI0731 and calculater.
What is this
Community fine-tunes of the CircleStone Anima base model (circlestone-labs/Anima-Base-v1.0-Diffusers),
converted into the WebGPU inference runtime Karume's container format (a .krm part
sequence whose first part carries the graph and model descriptors). Runs as-is in the
browser and in Deno.
- Ordinary many-step sampling β 20 steps / guidance 4 by default. Classifier-free guidance is on, which is what makes the negative prompt take effect. The official models (base / Aesthetic / Turbo) live in hdae/karume-anima.
- Not readable by diffusers (it's a different container with an embedded graph); the reader is a pipeline that implements
anima/1. - Exporter used for the conversion:
karume/0.23.0. The distribution manifest iskarume.json(karume/6).
Models and their origins
Each model below is a community fine-tune of the CircleStone Anima base model. The text
encoder, VAE and tokenizers are shared with the official repository
(hdae/karume-anima) through pinned
cross-repository references in karume.json.
anima-wai-v1.0 β WAI-ANIMA v1.0 (base 1.0)
- Author: WAI0731
- Source: https://civitai.com/models/2544636/wai-anima
- Converted from:
waiANIMA_v10Base10.safetensors
Permissions listed on the source page (as of 2026-08-22):
allowNoCredit: trueallowCommercialUse: Image / RentCivitallowDerivatives: trueallowDifferentLicense: true
anima-copycat-20260610 β copycat-anima 20260610
- Author: calculater
- Source: https://civitai.com/models/2377376/copycat-anima
- Converted from:
copycatAnima_20260610.safetensors
Permissions listed on the source page (as of 2026-08-22):
allowNoCredit: trueallowCommercialUse: Image / RentCivitallowDerivatives: trueallowDifferentLicense: false
License
Every model here derives from the CircleStone Anima base model and stays under the
CircleStone Labs Non-Commercial License v1.2 (non-commercial use only). This
repository ships LICENSE.md (the full text of that license, followed by the NVIDIA
Open Model License Agreement) and NOTICE.md (these attribution notices plus the list
of modifications).
Any rights to use the CircleStone Models and/or Derivatives in this repository are granted to you directly by CircleStone Labs LLC under the CircleStone Labs Non-Commercial License (LICENSE.md).
The CircleStone Model is licensed by CircleStone Labs LLC under the CircleStone Non-Commercial License. Copyright CircleStone Labs LLC. IN NO EVENT SHALL CIRCLESTONE LABS LLC BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
Licensed by NVIDIA Corporation under the NVIDIA Open Model License
The CircleStone Anima model is a Derivative Model of NVIDIA Cosmos-Predict2-2B-Text2Image (https://huggingface.co/nvidia/Cosmos-Predict2-2B-Text2Image), so the weights in this repository are also subject to the NVIDIA Open Model License Agreement (https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/) insofar as it applies to Derivative Models. A copy of that agreement is included in LICENSE.md. Both licenses apply; the NVIDIA agreement does not widen the CircleStone Non-Commercial License.
Built on NVIDIA Cosmos.
- Outputs you generate are yours to use for any purpose, including commercially (license Β§2(e)); the non-commercial restriction applies to the model weights and derivatives, not to outputs.
- The community fine-tunes are redistributed under the permissions their source pages state (listed per model above); those permissions do not widen the base model's license.
anima-copycat-20260610: the source page setsallowDifferentLicenseto false, so this redistribution keeps the same terms β do not relicense it.- This is not an official product of CircleStone Labs LLC, and it is not endorsed, approved or validated by CircleStone Labs LLC.
Models
| Model | Pipeline | Quants | Default quant |
|---|---|---|---|
anima-wai-v1.0 (default) |
anima/1 |
bf16 / bf16+dit8 / bf16+dit8-a8 / bf16+dit8-a8-attn8 / bf16+dit8-a8-attn8-s16 / bf16-c16 |
bf16+dit8-a8-attn8-s16 |
anima-copycat-20260610 |
anima/1 |
bf16 / bf16+dit8 / bf16+dit8-a8 / bf16+dit8-a8-attn8 / bf16+dit8-a8-attn8-s16 / bf16-c16 |
bf16+dit8-a8-attn8-s16 |
model selects one of these; omitted, it is anima-wai-v1.0. quant defaults to that model's own default quant.
Usage
import { encodePng } from "jsr:@karume/models";
import { Pipeline } from "jsr:@karume/models/anima";
await using pipeline = await Pipeline.fromPretrained({
repo: "hdae/karume-anima-extra",
// Pin a commit for reproducible builds β without it you track `main`, and a future
// repo update (renamed files, new manifest format) may break your app.
// Copy the full hash from this repo's "Files and versions" tab:
// revision: "<full commit sha>",
}, {
// model: "anima-wai-v1.0", // default β available: anima-copycat-20260610 / anima-wai-v1.0
// quant: "bf16+dit8-a8-attn8-s16", // default β available: bf16 / bf16+dit8 / bf16+dit8-a8 / bf16+dit8-a8-attn8 / bf16+dit8-a8-attn8-s16 / bf16-c16
});
const image = await pipeline.generate({
prompt: "1girl, solo, long hair, blue eyes, school uniform, masterpiece",
// steps: 20, // default β more steps trade time for detail
// Resolution β non-square is fine; each side on a 16 px grid, between 512 and 2048 px:
// Sides above 1920 px sit outside the range this model declares (its RoPE tables
// declare 120 tokens = 1920 px per side); those positions are extrapolated with the
// same formula, so what can differ there is the image quality.
// resolution: { width: 1024, height: 1024 }, // default
// Classifier-free guidance runs a second (uncond) branch β twice the work per step.
// It is on by default here, which is what makes the negative prompt take effect:
// guidanceScale: 4, // default
// negativePrompt: "low quality, worst quality, blurry, bad anatomy, jpeg artifacts", // default
seed: 42, // same seed + same request β same image
});
const png = await encodePng(image.data, image.width, image.height);
await Deno.writeFile("anima.png", png);
Weights are fetched once and cached (verified against karume.json's size / sha256).
Model: anima-wai-v1.0
Quants
| Quant | What it is | Download | Weights | Compute |
|---|---|---|---|---|
bf16 |
Original weights (bf16) β All four components stored as bf16 with f32 compute β the upstream checkpoint's own values, bit for bit, and the largest download. | 5.06 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = bf16 / vae_decoder = bf16 |
β |
bf16+dit8 |
int8 transformer β Transformer stored as int8 (half its bf16 size) and computed in f32; the text encoder, text conditioner and VAE decoder stay bf16. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
β |
bf16+dit8-a8 |
int8 transformer, int8 linear β The int8 transformer with per-token int8 activations in its linear layers, multiplied as integers (dp4a where the GPU has it); the same files as bf16+dit8. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
linearCompute = a8 |
bf16+dit8-a8-attn8 |
int8 transformer, int8 linear and attention β As bf16+dit8-a8, with int8 activations inside attention too; the same files as bf16+dit8. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
linearCompute = a8 / attentionCompute = a8 |
bf16+dit8-a8-attn8-s16 (default) |
int8 transformer, int8 linear and attention, f16 scores β As bf16+dit8-a8-attn8, with the attention scores held in f16; the same files as bf16+dit8. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
linearCompute = a8 / attentionCompute = a8 / attentionScoreStorage = f16 |
bf16-c16 |
Original weights, f16 transformer compute β The bf16 preset's files, with the transformer's linear layers and attention computed in f16; the text encoder, text conditioner and VAE decoder compute in f32. Needs the shader-f16 GPU feature. | 5.06 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = bf16 / vae_decoder = bf16 |
linearCompute = f16 / attentionCompute = f16 / requires shaderF16 |
If no quant is given, it runs as bf16+dit8-a8-attn8-s16 (this model's recommended default).
In a quant name, dit is the transformer component.
Per-file size and sha256 live in karume.json β verify against that at the fetch layer.
Dtype labels use the runtime's storage dtype vocabulary (f16 / bf16 / i8 / i4 / i2), not the fp16 spelling common elsewhere in the ecosystem.
Weights ship as Karume container files (.krm), split into numbered parts; a part is fetched and verified on its own.
A path under shared/ is one this model shares byte for byte with another model in this repository (it is fetched and cached once).
Some components are fetched from hdae/karume-anima at commit 58be43f57ffbb6f2β¦ β those bytes are identical to this model's own, so they are not stored here a second time.
Defaults
Any knob not passed to generate() is filled in from the manifest's defaults.
- steps: 20
- guidanceScale: 4
- resolution: 1024 Γ 1024
- negativePrompt:
low quality, worst quality, blurry, bad anatomy, jpeg artifacts
Model: anima-copycat-20260610
Quants
| Quant | What it is | Download | Weights | Compute |
|---|---|---|---|---|
bf16 |
Original weights (bf16) β All four components stored as bf16 with f32 compute β the upstream checkpoint's own values, bit for bit, and the largest download. | 5.06 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = bf16 / vae_decoder = bf16 |
β |
bf16+dit8 |
int8 transformer β Transformer stored as int8 (half its bf16 size) and computed in f32; the text encoder, text conditioner and VAE decoder stay bf16. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
β |
bf16+dit8-a8 |
int8 transformer, int8 linear β The int8 transformer with per-token int8 activations in its linear layers, multiplied as integers (dp4a where the GPU has it); the same files as bf16+dit8. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
linearCompute = a8 |
bf16+dit8-a8-attn8 |
int8 transformer, int8 linear and attention β As bf16+dit8-a8, with int8 activations inside attention too; the same files as bf16+dit8. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
linearCompute = a8 / attentionCompute = a8 |
bf16+dit8-a8-attn8-s16 (default) |
int8 transformer, int8 linear and attention, f16 scores β As bf16+dit8-a8-attn8, with the attention scores held in f16; the same files as bf16+dit8. | 3.24 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = i8 / vae_decoder = bf16 |
linearCompute = a8 / attentionCompute = a8 / attentionScoreStorage = f16 |
bf16-c16 |
Original weights, f16 transformer compute β The bf16 preset's files, with the transformer's linear layers and attention computed in f16; the text encoder, text conditioner and VAE decoder compute in f32. Needs the shader-f16 GPU feature. | 5.06 GiB (1.16 GiB shared) | text_encoder = bf16 / text_conditioner = bf16 / transformer = bf16 / vae_decoder = bf16 |
linearCompute = f16 / attentionCompute = f16 / requires shaderF16 |
If no quant is given, it runs as bf16+dit8-a8-attn8-s16 (this model's recommended default).
In a quant name, dit is the transformer component.
Per-file size and sha256 live in karume.json β verify against that at the fetch layer.
Dtype labels use the runtime's storage dtype vocabulary (f16 / bf16 / i8 / i4 / i2), not the fp16 spelling common elsewhere in the ecosystem.
Weights ship as Karume container files (.krm), split into numbered parts; a part is fetched and verified on its own.
A path under shared/ is one this model shares byte for byte with another model in this repository (it is fetched and cached once).
Some components are fetched from hdae/karume-anima at commit 58be43f57ffbb6f2β¦ β those bytes are identical to this model's own, so they are not stored here a second time.
Defaults
Any knob not passed to generate() is filled in from the manifest's defaults.
- steps: 20
- guidanceScale: 4
- resolution: 1024 Γ 1024
- negativePrompt:
low quality, worst quality, blurry, bad anatomy, jpeg artifacts
Model tree for hdae/karume-anima-extra
Base model
nvidia/Cosmos-Predict2-2B-Text2Image