This workflow is used to get a starting point when creating captions for a character LoRA. You feed it an image and it will analyze the image and provide a well rounded description aimed for Character LoRA that should include all the relevant details required for proper captioning.I found a good Ollama model with vision capability that isn't censored and fed it with a detailed prompt on how to craft the proper caption for a Character LoRA. The Ollama model is about 7GB and should fit on almost any GPU if ran on its own.An additional "hint" box is added, so the user can provide a hint for the Vision model for exceptional situation like extreme close-up where the image analysis can get wrong if the AI isn't provided with at least minimal context. Leave that box empty in most cases.I have also added a few markdown notes with tips for the best possible captioning.

LoRA Dataset Caption Aid with Ollama Vision Uncensored

Initial Version.  I indicated "Flux" as the model but it can work with any model - it's just a tool for help with captioning your LoRA dataset, it will work for Flux Dev, Flux Krea, WAN, Qwen Image, etc.

Files with SHA256: dec49a5f93b31c337aa82cd6893bab198469233c51b365759bc5bafe4cfecb05