Lanzi gen-ai-pose-image 0.3.0 (DSH pose image generation plugin, adds direct browser connection to ComfyUI and a face reference library)

DeepSeek Harness plugin and standalone web studio: Select poses from a public pose line-art library and outfits from a person-free clothing image library to generate pose-change / outfit-change images of the same model (default is the community whale-girl character "Big Fat Fish", but you can also import photos of other models). 10-31 featured it; 11-01 later published 0.3.0 on npm.

DeepSeek Harness plugin and standalone web studio: Select poses from a public pose line-art library and outfits from a person-free clothing image library to generate pose-change / outfit-change images of the same model (default is the community whale-girl character "Big Fat Fish", but you can also import photos of other models). 10-31 featured it; 11-01 later published 0.3.0 on npm.

Star count 290

Reasons

  • 0.3.0 adds direct browser connection to ComfyUI: In the online web "Image Generation Settings", select ComfyUI and enter the local or remote HTTPS address. The browser directly uploads, submits tasks, and retrieves results; the website does not proxy requests, nor does it require starting the website on the GPU computer. Includes a set of workflows with toggles (outfit change, pose change, both, up to a four-panel collage).
  • 0.3.0 adds face references: You can select one image from the public CC0 model face database. Facial features, expressions, and hairstyles follow this image, while poses and outfits still follow their respective references. The npm package includes a face catalog file, filterable by face close-up / style and hair length.
  • Original cloud image generation is retained: Supports Volcano Ark Seedream, Gemini, and OpenAI-compatible interfaces; if no key is provided, it only outputs the prompt for each image, which can be handed to dsh-image-gen for generation; the proxy side has four tools: search_poses, search_costumes, search_faces, and generate_pose_images.

Usage

  • Install into DSH: dsh plugin --profile web add gen-ai-pose-image. After restarting the profile, "Lanzi / Studio" appears in the left sidebar. For standalone execution, Node.js 22.19+ is required. Run npm install && npm run dev, then open http://127.0.0.1:43123。
  • Connect to local ComfyUI: In the web interface, add --enable-cors-header <website origin> to the http://127.0.0.1:8188,并在 ComfyUI startup command. For phones or other computers, enter the HTTPS entry point of the GPU computer. The documentation provides a Caddy reverse proxy example with origin validation and Bearer tokens. Workflow files are located in shared/workflows/.

Risks

  • Asset copyrights are scattered: Upstream pose line-art notes prohibit commercial use. The "Big Fat Fish" character design is under CC BY-NC-SA 4.0. Only the code in the repository is MIT licensed. GitHub cannot identify the overall license (NOASSERTION). Verify each item individually before commercial use.
  • When connecting remotely to ComfyUI, native ComfyUI does not automatically become protected just because a token is entered. The entry point must handle authentication itself. The documentation explicitly warns against exposing only the 8188 port. The effect images in the README have not been verified; image generation quality depends on the connected model and workflow.

Open

GitHub

Explore the latest AI Radar updates