# LLaVa Img2Txt From this reddit post: https://www.reddit.com/r/Oobabooga/comments/12xaktr/llava_support_has_been_added/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button Downloaded: https://huggingface.co/wojtab/llava-13b-v0-4bit-128g/blob/main/llava-13b-v0-4bit-128g.safetensors Don't forget the json... PR: https://github.com/oobabooga/text-generation-webui/pull/1487 README: https://github.com/oobabooga/text-generation-webui/blob/main/extensions/llava/README.md copied start-webui-vicuna-gpu and edited to add: ``` call python server.py --wbits 4 --model llava-13b-v0-4bit-128g --groupsize 128 --model_type LLaMa --extensions llava --extensions api --listen --auto-devices --chat ``` I then had to go to setting and enable the "llava" checkbox and restart the UI. It's working! ![](https://i.imgur.com/eNMvfmc.png) I'm trying to figure out how to get it to give a stablediffusion style output... ``` Provide a description of this image, but I want you to do it in a specific fashion that's used for stablediffusion, an AI that generates images for text2image. In this format you describe portions of the image in short phrases (1-4 words) separated by commas. First start with the subject, then, details about the style and the portions of the image. Then add artists who are similar, art styles and schools that match the image, and websites or periodicals that have imagery similar to the image. To emphasize a particular phrase, surround it in parenthesis, the more parenthesis around a phrase or word, the more emphasis will be applied. Here are some examples to follow for the description: -- sun goddess, exquisitely detailed, detailed face, digital painting, artstation, standing, concept art, soft, sharp focus, illustration, art by artgerm and alphonse mucha, high definition digital art, magical lighting, trending on artstation, sharp focus, studio photo, intricate details, highly detailed, by greg rutkowski -- private detective from the 50's, (((digital painting))), detailed gorgeous face, perfect anatomy, centered, chiaroscuro, (dark atmosphere), action pose, smooth, art by Carne Griffiths and Wadim Kashin, trending on artstation, sharp focus, intricate details, highly detailed, kids story book style, watercolor style, -- mixpnk style, female private detective from the 50's, (((digital painting))), detailed gorgeous face, perfect anatomy, centered, chiaroscuro, (dark atmosphere), action pose, smooth, art by Carne Griffiths and Wadim Kashin, trending on artstation, sharp focus, intricate details, highly detailed, kids story book style, watercolor style, ```