Description This is an Anima workflow that generates high-quality Silly Tavern character expressions in just minutes. Gives you tons of options (portrait mode, background creator, low resolution, manual selection, additional prompt tags, etc.), yet, it's designed to be easy as just clicking "run" upon first load. In addition, it will generate sex expressions for all non-extreme positions, as well as combat poses. (Currently prototype mode, meaning this is highly unrefined, so expect a lot of updates) Update 0.7 Notes: Quick stuff: Tag updates, added clothing toggles, removed reference image, Remove background now handled by Qwen Image 2.1 (No more pixel color picker and REMBG), improved comic line handling (still WIP). As I work on 0.8, I realize tags have been greatly neglected, so concentration on that. Also, I'll probably remove silly sex positions and put foot job in place of something like amazon. Quick Note: Just found this out after posting. Qwen Img slows down generation so much, that the counters at the bottom don't work as well when using non-random generation. Temp fix: Click on each one and set the timer from 30 seconds to 90 seconds. I'll fix this on the next go-around. Pre-Requirements:<ul><li>[ <a target="_blank" rel="ugc" href="https://github.com/Comfy-Org/ComfyUI-Manager">Download</a> ]: ComfyUI Manager (It allows you to install missing nodes) (Or: pip install -U --pre comfyui-manager)</li><li>[ <a target="_blank" rel="ugc" href="https://github.com/yolain/Comfyui-Easy-Use">Download</a> ]: Easy-Use Nodes (For wildcards and style selector) (It's in the ComfyUI Manager repo too - But it's the main nodes I use)</li><li>[ <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/tree/main">Download</a> ]: Qwen Image 2.1 Model (Either the int 8 safe tensor here: [ <a target="_blank" rel="ugc" href="https://huggingface.co/Abiray/Qwen-Image-2.1-GGUF/tree/main">Download</a> ] ) (Or, if you have low VRAM, use the GGUF here: [ <a target="_blank" rel="ugc" href="https://huggingface.co/Abiray/Qwen-Image-2.1-GGUF/tree/main">Download</a> ] ) </li></ul>Operation<ul><li>Open ComfyUI, press 'W' for Workflows, click on ST_Expression_Gen </li><li>Then select a character from the Character Styles node. </li><li>Click 'Run' (Or batch run) </li><li>It will generate a random expression in this format: expression-##.avif (As well as a JSON containing the prompt and the seed) </li><li>Copy all avif files to SillyTavern/data/default-user/character/&lt;character&gt; </li><li>Restart SillyTavern, it now should recognize your character expressions </li></ul>Making your own character: <ul><li>Go to ComfyUI/custom_nodes/comfyui-easy-use/styles </li><li>Open the Characters.json (Or better yet, make your own, so your characters don't get overwritten on update) (Use Characters.json as a template.) (Use only character tags) (No quality tags, no action tags, gestures, posture, poses, camera, etc..) (Only use anatomy and clothing tags.)</li></ul>Since this is in prototype mode, you'll probably have to curate about 50% of the images and delete the ones that don't turn out well. QuestionsIs it perfect? No, but it's output will outdo 99% of other expressions out there. And, I have done my best to mitigate any limitations dealing with auto generation expressions. It's also in prototype phase, there are some wrinkles to iron out. What model should I use? Start with WAI - <a target="_blank" rel="ugc" href="https://civarchive.com/models/2544636/wai-anima">https://civarchive.com/models/2544636/wai-anima</a> How do I get Silly Tavern to use the extra expressions? Set expressions to Main API. Add custom expression for each one you want. (You only have to do this once.)

SillyTavern Auto Expressions

<ul><li>Tag updates, fixes, and tweaks</li><li>Qwen Image 2.1 - No more Pixel Color Picker / REMBG</li><li>Better Comic Outline Algorithm</li><li>Remove reference image (Just didn't work well, when implemented)</li><li>Removed Mid Frame output for Minimax</li><li>Image prompt output tweaks and fixes</li><li>Added Clothing / Footwear / Headwear style sheets</li></ul>

Files with SHA256: 373ce048b6f004723b7fefda1773b350965f412336520340568bca9b917f5523