<h1 id="instant-lora-q310xz7gr">Instant LoRA</h1><h2 id="legacy-lore-(for-the-original-revamp-version):-4z2li8o3j">Legacy Lore (for the Original Revamp Version):</h2><h2 id="original-concept-by-aloevera_purple(no-longer-on-civitai-good-luck-wherever-the-winds-of-change-may-take-you!)-r63akpk0u">Original Concept by AloeVera_Purple (No longer on CivitAI, good luck wherever the winds of change may take you!)</h2>In the words of its original creator: "This workflow will take your input images, distribute them into the two different IPAdapter models, [Edit: Only one for Flux.1 because there is no Face IPAdapter yet that works on Flux.] with different timesteps (distribution in time basically), and shift your results, like a LORA would... but in this case, you don't have to train anything. It has likeness scores up to 95%. + With the right settings, it can be incredibly accurate. Unfortunately, there are no "Golden" settings that work for all. You'll have to tweak things a little bit, up and down. See what works best for you."<h2 id="information:-ybiglnqwq">Information:</h2>From now on all information pertaining to the versions can be found in the version history section at the right side of the screen (Just below the download section). This will contain a list of changes and/or Installation Steps for the Workflow and the Files (If you downloaded those).<h2 id="how-to-use:-ae8fbqko9">How to use:</h2>The general usage of the workflow is pretty simple. Just add up to 6 images of your choice to the Image Loaders (Bypass the ones you want to leave empty), type up your prompt and send it to "Generate". The IPAdapter will combine the Images into a makeshift LoRA and use the style and overall color scheme on the new generation.<ul><li>The Batch Maker Node scales with additional Images so feel free to add as many as you want. Keep in mind that more Images means longer wait time for IPAdapter to process the images into the Makeshift LoRA.</li></ul><h3 id="works-in-progress-and-suggestions-kuphd345s">Works in Progress and Suggestions</h3><ul><li>Work is being done to add Regional Prompting to the Workflow. At the moment I'm figuring out how to do add this function into my "Negative Prompt &amp; HiRes.Fix" workflow. once this all works accordingly it'll be added to the Instant LoRA workflow as well!<ul><li>Found Issue: Tying in the Regional Prompting is not a problem but it seems the IPadapter for Flux doesn't like playing together with it. The Workflow seems to work (no errors during execution of the workflow) but the output doesn't adhere to the regional mask prompts. Will investigate further. There are no errors thus in theory it should work.</li></ul></li><li>Adding a Weight or Ratio system to the Images so you can choose how much it'll influence the generated picture. ~ Suggested by <a target="_blank" rel="ugc" href="https://civarchive.com/user/angrysky">angrysky</a><ul><li>Found Issue: While it is possible to tie weight markers to each image this will at the moment require the Batch Loader to be turned into a Model &gt; Model Feed system and for Flux.1 this will result into an extreme long generation time even on higher end Consumer GPUs. (I will do some more research but at the moment while possible the extended generation time makes this practically non-viable.)</li></ul></li></ul>

Instant LoRA

Changes:<ul><li>Finally there is Flux.1 Functionality.<ul><li>I have chosen for the GGUF version of Flux.1 so more people can play around with the Workflow as each can choose their own quantized version that works best for him/her.<ul><li>The Quantized GGUF files can be found at:<ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/city96/FLUX.1-dev-gguf/tree/main">https://huggingface.co/city96/FLUX.1-dev-gguf/tree/main</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/tree/main">https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/tree/main</a></li></ul></li><li>Chose whichever filesize will work for your rig.</li></ul></li></ul></li></ul>Setup:<ul><li>Install ComfyUI Manager by ltdrdata ( <a target="_blank" rel="ugc" href="https://github.com/ltdrdata/ComfyUI-Manager">https://github.com/ltdrdata/ComfyUI-Manager</a> )<ul><li>This will make sure you can easily install the missing nodes you'll need for this Workflow to function.</li></ul></li><li>Download the *.zip file containing the Workflow.<ul><li>A separate Download with all the needed files is also available.</li></ul></li><li>Launch ComfyUI and Load the Workflow.</li><li>Use Manager to install the Missing Nodes and then restart ComfyUI (there should be no Red Colored Nodes after this step)</li><li>Put the Extra Files into the Correct Folders<ul><li>CLIP-ViT-Large-patch14.safetensors --&gt; Copy this into your ComfyUI\models\clip_vision Folder</li><li>flux-ip-adapter-v2.safetensors --&gt; Copy this into your ComfyUI\xlabs\ipadapters Folder<ul><li>If at this point you don't have these folders in the Model folder it's perfectly okay to create them.</li></ul></li><li>The 2 CLIP files need to go into the ComfyUI\models\clip Folder<ul><li>I highly suggest to use the LongCLIP-L file if your CPU and GPU can handle it for better results. If not the normal clip_l is perfectly acceptable.</li></ul></li><li>I have Provided 4 upscale_models in the *.zip file these should go into the ComfyUI\models\upscale_models Folder.</li></ul></li></ul>

Files with SHA256: 5ceea8e98c1345ff82ae91c1cc41afda7d57de10732655e4b2b07fa22a9d91e0