<h3 id="comfyui-workflow-for-image-generation">ComfyUI Workflow for Image Generation</h3><ul><li>This workflow is not compatible with RvTools_v2 !</li><li>it requires the new Version of RvTools, which is also the one that is used from now on for my Workflows, but i made sure that both versions don't conflict each other.</li></ul><ul><li>Needed Custom Node:<ul><li>Git-Hub: <a target="_blank" rel="ugc" href="https://github.com/r-vage/ComfyUI-RvTools-X">RvTools-X v2.5.0</a></li><li>Comfyui-Registry: <a target="_blank" rel="ugc" href="https://registry.comfy.org/publishers/rvage/nodes/ComfyUI-RvTools-X">RvTools-X v2.5.0</a>if you install the node with the manager choose nightly because i don't upload every change to comfyui registry. if you choose latest you'll install the latest version stored on comfy reg. nightly is the current version from git hub.</li></ul></li><li>if you get type mismatch errors, install the frontend version 1.26.0<ul><li>python_embeded\python.exe -m pip install comfyui-frontend-package==1.26.0</li></ul></li><li>if you want to use pulid or reactor you need to install numpy 1.26.4<ul><li>python_embeded\python.exe -m pip install numpy==1.26.4</li></ul></li><li><a target="_blank" rel="ugc" href="https://civarchive.com/articles/15787/listing-links-resources">Resource Listing</a></li><li>when using torch 2.7.x+cu128 use triton 3.3.1 (3.4 will cause issues)<ul><li>python_embeded\python.exe -m pip install -U "triton-windows==3.3.1.post19"</li></ul></li></ul>When you look at the image showing the workflow, you can see that it is divided into groups. Each group offers useful tools for influencing the image to be generated. Almost every group has a ctx (context menu) for input and one at the end for output of this group. If there are groups in the workflow that you do not need or cannot use, select the entire group (hold Ctrl + draw a rectangle with the left mouse button over the area) with all nodes in it and the two ctx and delete them. The same applies if you want to duplicate a group e.g. a detailer. Select the group+ctx, copy it, and paste it into a free space. Connect the two ctx with the ones where the group is to be inserted.the workflow has 2 stop funtions:<ul><li>one is in the initial render group, it allows you, when using a random seed, to stop after each render and choose if you want to quene for a new image or set the seed to fixed and process this image.</li><li>the 2nd is before the wan groups. if this one is enabled and auto queue is also enabled the wan groups are ignored. means you can render multible images without creating a video of each of them. if auto queue is not enabled the workflow stops after the image is processed and you can adjust the prompt for the video in the wan: prompt group, disable the stop option after that and start a new queue (fixed seed!) to create a video of the generated image.</li></ul>

iGEN II-X for FLUX | SD | HiDream | Qwen (STD | GGUF)

<ul><li>this workflow requires RvTools-X, it is NOT compatible with RvTools_v2 but both can run side by side.</li><li>if you have downloaded this workflow already you have to download RvTools-X again. i had to do a lot of code cleaning so i deleted the old version. (i didnt want to fill the history will bullshyte^^)</li><li>reuploaded with an update: ive added stop buttons in the detailers (with on/off switch) i noticed a lot of fail detections. this helps to adjust the seed, check the detection output and continue if its ok, same for sam2 output to test the sam2 settings i.e mask grow/blur, process_detail. detail_erode/dilate etc.)</li><li>updated, fixed and rebuilded from version 2.1</li><li>fixed the llm prompt, it is created and saved into the image but not actually used to create the image, in the old version.</li><li>added a second seed node, one for the image and one for the prompt. this allows you to stop the random prompts. if you have an image that you like, set the prompt seed to fixed and the image seed to random to create images with that prompt.</li><li>fixed the supir subgraph, extracted the lora from it.</li><li>i've removed some nodes because rvt-x don't need them anymore, i.e. save images can handle direct connections from the context menu now, also without image input if the output image is transported within the context menu. it also allows you to use placeholders like %basemodel etc.</li><li>much more small adjustments like this.</li></ul>

Files with SHA256: f4297cae578c55e164821cd36951624afe7fa9afca57b4fed5d3b0103a263242