Problem Statement
Jan is very plain in the models, and there is no adequate user friendly ui for sd.cpp, so why not enhance the experience with adding sd.cpp?
Feature Idea
The llms that can toolcall can call 4 more but built in offline non mcp provided tools, image_gen, image_edit, video_gen, video_i2v, images have internal links that models can see and can be used for different things, like telling it to edit 2 already generated images into one using image_edit, or use the generated/user send image as the start for video_i2v. Importing window for models has choosable tags like "Model" "Diffusion model" (used for z-image-turbo) "Llm" some of the "Clip_X" and "t5xxl" (like gguf import window for llms but not only mmproj, you can select the type of gguf weights in a dropout), with also model type, and have all other sd.cpp stuff in the settings for it, also an option if the llm an see the output (image only)
Problem Statement
Jan is very plain in the models, and there is no adequate user friendly ui for sd.cpp, so why not enhance the experience with adding sd.cpp?
Feature Idea
The llms that can toolcall can call 4 more but built in offline non mcp provided tools, image_gen, image_edit, video_gen, video_i2v, images have internal links that models can see and can be used for different things, like telling it to edit 2 already generated images into one using image_edit, or use the generated/user send image as the start for video_i2v. Importing window for models has choosable tags like "Model" "Diffusion model" (used for z-image-turbo) "Llm" some of the "Clip_X" and "t5xxl" (like gguf import window for llms but not only mmproj, you can select the type of gguf weights in a dropout), with also model type, and have all other sd.cpp stuff in the settings for it, also an option if the llm an see the output (image only)