<h1 id="hooch-anima-3.8b-refined-fusion">Hooch Anima 3.8B — Refined Fusion</h1>A new branch of the Hooch Anima family, built on the 52-block Anima 3.8B architecture.Hooch Anima 3.8B is an experimental checkpoint merge.The resulting model has been structurally verified:<ul><li>Architecture: Anima 3.8B — 52 transformer blocks (0–51)</li><li>Precision: BF16</li><li>File size: approximately 7.5 GB</li><li>Remapping: Anima 2.9B (40 blocks) → Anima 3.8B (52 blocks)</li><li>Semantic Connector: not embedded; external sidecar used for the recommended workflow</li></ul><h2 id="visual-style">Visual style</h2>Hooch Anima 3.8B focuses on expressive anime characters, detailed fantasy clothing and armor, vibrant illustrations and character-focused compositions and NSFW.Depending on the prompt, it can produce cleaner anime-style artwork or richer, more dimensional fantasy illustrations.It is particularly suited to fantasy characters, portraits, adventurers, magical settings and dramatic illustrations.This is an experimental merge, so different prompts and seeds may produce noticeably different stylistic results.<h2 id="recommended-generation-settings">Recommended generation settings</h2>These settings were selected based on local testing.Sampler: Euler a (Er sde)Scheduler: NormalSteps: 30CFG: 6 (range 5-6)Resolution Approximately 1 megapixelOther settings may also work, but the configuration above is the recommended starting point.<h2 id="comfyui-compatibility">ComfyUI compatibility</h2><h3 id="standard-anima-workflow-supported">Standard Anima workflow — supported</h3>Hooch Anima 3.8B can also generate images using the standard Anima workflow with the native Qwen3 0.6B text encoder.This is the simpler setup and does not require the external Semantic Connector.However, images may differ considerably from those generated with the dedicated 3.8B workflow.<h3 id="anima-3.8b-q35-workflow-recommended">Anima 3.8B Q35 workflow — recommended</h3>For the best results observed during testing, I recommend the dedicated Anima 3.8B workflow using Qwen3.5 4B conditioning and the external Semantic Connector.This is the configuration used for the recommended showcase results.Required additional files:<ul><li><code>anima_connector_only.safetensors</code> — external Semantic Connector sidecar</li><li><code>qwen35_4b.safetensors</code> — Qwen3.5 4B text encoder</li><li><code>qwen_3_06b_base.safetensors</code> — native Anima text encoder</li><li><code>qwen_image_vae.safetensors</code> — VAE</li></ul>Place the two text encoders inside <code>ComfyUI/models/text_encoders/</code>, the VAE inside <code>ComfyUI/models/vae/</code>, and the connector sidecar alongside the diffusion checkpoint.Custom nodes: the recommended workflow uses the Anima 3.8B runtime and the Delta Q35 V2 extension. Install the versions compatible with the supplied workflow.The external connector is not a LoRA and should not be loaded through a LoRA Loader.<h2 id="downloads-and-dependencies">Downloads and dependencies</h2>Main checkpoint: Hooch Anima 3.8B BF16 — available through this model page.Qwen3.5 4B text encoder<a target="_blank" rel="ugc" href="https://huggingface.co/lylogummy/Anima-3.8B/tree/main/text_encoders">https://huggingface.co/lylogummy/Anima-3.8B/tree/main/text_encoders</a>Download <code>qwen35_4b.safetensors</code>.Native Anima Qwen3 0.6B text encoder<a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima/tree/main/split_files/text_encoders">https://huggingface.co/circlestone-labs/Anima/tree/main/split_files/text_encoders</a>Download <code>qwen_3_06b_base.safetensors</code>.Anima VAE<a target="_blank" rel="ugc" href="https://huggingface.co/circlestone-labs/Anima/tree/main/split_files/vae">https://huggingface.co/circlestone-labs/Anima/tree/main/split_files/vae</a>Download <code>qwen_image_vae.safetensors</code>.External Semantic Connector<a target="_blank" rel="ugc" href="https://civarchive.com/models/2913466">https://civarchive.com/models/2913466</a>Find <code>anima_connector_only.safetensors</code> among the additional or optional files.Anima 3.8B custom nodes<a target="_blank" rel="ugc" href="https://github.com/GumGum10/comfyui-anima-3-8B">https://github.com/GumGum10/comfyui-anima-3-8B</a>Delta Q35 V2 custom nodes<a target="_blank" rel="ugc" href="https://github.com/bestluner-create/ComfyUI-Anima-Delta-Q35v2">https://github.com/bestluner-create/ComfyUI-Anima-Delta-Q35v2</a><h2 id="workflow">Workflow</h2>The recommended ComfyUI workflow uses the external connector and Qwen3.5 conditioning.Use the supplied JSON workflow, if attached, to reproduce the recommended setup. Make sure all required nodes and model files are installed before loading it.For users who prefer a simpler configuration, the standard Anima workflow is also supported.<h2 id="license">License</h2>This checkpoint is derived from the Anima model family and follows the applicable upstream model licenses, including non-commercial restrictions.The model weights must not be treated as unrestricted commercial-use assets.Generated images are governed separately by the applicable license terms.Enjoy experimenting with Hooch Anima 3.8B! And if you like it please share your creation!!!

Hooch Anima 3.8B

Hooch Anima 3.8B is a new merge, preserving the full 52-block Anima 3.8B architecture in BF16.A blend of expressive anime aesthetics, detailed fantasy artwork and the distinctive visual character of Hooch.

Files with SHA256: ca33f4e4582b8d68faa3af83275c563fd76dcd9a3219d2e95c380949413f1fae