To-Do

[daniandtheweb]

This issue will serve as the master to-do list for sd.cpp-webui. Feel free to tackle any of the items below, or suggest new ones by commenting!

Core Features

• Implement the missing features from stable-diffusion.cpp.
• Job queue.
• Fix photomaker.
• Qwen Image Edit and Flux Kontext support.

Frontend & UI/UX

• Redesign the UI for better user experience.
• Add a theming option.

Documentation & Testing

• Improve the README.md.
• Implement unit tests.

[phil2sat]

qwen image edit and flux Kontext fails on three things.

"--init-img" has to be "-r" for qwen image-edit
"--strength None" in command makes sd fail to run
"-M img_gen" needs to go
dont know why where what strength none is for as from the img2img there is no possibility to change this parameter

after this it should be possible to use the img2img page

problem is to check for EDIT model:

strings qwen-rapid-aio-nsfw-v5-Q3_K_S.gguf | grep -i isdedit       
IsdEdit~        X

But needs to get killed and parser consumes ressources for every model in folder (BAD).
Better to use a CheckBox beneath unet "is Kontext/Edit Model" then remove "--strenght", replace "--init-img" to "-r" and remove "-M img_gen" for both
Tested so far only for edit but kontext should be the same from documentation. but will test...

Failing:
./sd -M img_gen -p "Change test to..." --sampling-method euler_a --steps 10 --scheduler discrete -W 832 -H 1216 -b 1 --cfg-scale 2.5 -s -1 --clip-skip -1 --embd-dir /home/phil2sat/sd.cpp-webui/models/embeddings/ --lora-model-dir /home/phil2sat/sd.cpp-webui/models/loras/ -t 8 --rng cuda -o /home/phil2sat/sd.cpp-webui/outputs/txt2img/193.png --diffusion-model /home/phil2sat/sd.cpp-webui/models/unet/qwen-rapid-aio-nsfw-v5-Q4_0.gguf --vae /home/phil2sat/sd.cpp-webui/models/vae/qwen_image_vae.safetensors --qwen2vl /home/phil2sat/sd.cpp-webui/models/clip/Qwen2.5-VL-7B-Instruct-abliterated.Q8_0.gguf --offload-to-cpu --vae-tiling --color -v --init-img /tmp/gradio/f8b48cf061ccee7dd7913438af17994d99571d14dc3b09343800868292005dc2/ComfyUI_00113_.png --strength None

Working:
sd -p "change text to..." -W 832 -H 1216 -t 8 --rng cuda -o /home/phil2sat/sd.cpp-webui/outputs/txt2img/194.png --diffusion-model /home/phil2sat/sd.cpp-webui/models/unet/qwen-rapid-aio-nsfw-v5-Q3_K_S.gguf --vae /home/phil2sat/sd.cpp-webui/models/vae/qwen_image_vae.safetensors --qwen2vl /home/phil2sat/sd.cpp-webui/models/clip/Qwen2.5-VL-7B-Instruct-abliterated.Q8_0.gguf --offload-to-cpu --vae-tiling --color -v -r /tmp/gradio/f8b48cf061ccee7dd7913438af17994d99571d14dc3b09343800868292005dc2/ComfyUI_00113_.png --cfg-scale 2.5 --sampling-method euler

[daniandtheweb]

Qwen image edit is not explicitly supported for now, as well as flux kontext. I need to slightly reorder the interface before enabling them. Maybe I'll just add another tab for image edit models as it's technically neither text to image or image to image.

[daniandtheweb]

Qwen Image Edit and Flux Kontext support has finally been added.