David Hendrickson(@TeksEdge):Here's a ~6GB local vision model that can caption/tag photos that cloud AI may refuse to describe. It’s a Qwen3.5-9B vision fine-tune built for detailed image captioning, including explicit adult content. So yes, that can mean finding/tagging images containing specific content without uploading your photo library anywhere. 💾 Q4_K_M → 5.63GB 👁️ vision projector → ~600MB 🦙 llama.cpp 🖥️ LM Studio 🦙 Ollama And in one independent local-VLM comparison, it caught small details that several larger 7B–27B models missed. That makes it interesting for 📸 photo search 🏷️ automatic tagging 🎨 LoRA dataset captioning 🗂️ offline image cataloging 🔞 NSFW content Your photos stay on your machine. 🔗 Link in ALT.

2026.10.10 16:00

Here's a ~6GB local vision model that can caption/tag photos that cloud AI may refuse to describe. It’s a Qwen3.5-9B vision fine-tune built for detailed image captioning, including explicit adult content. So yes, that can mean finding/tagging images containing specific content without uploading your photo library anywhere. 💾 Q4_K_M → 5.63GB 👁️ vision projector → ~600MB 🦙 llama.cpp 🖥️ LM Studio 🦙 Ollama And in one independent local-VLM comparison, it caught small details that several larger 7B–27B models missed. That makes it interesting for 📸 photo search 🏷️ automatic tagging 🎨 LoRA dataset captioning 🗂️ offline image cataloging 🔞 NSFW content Your photos stay on your machine. 🔗 Link in ALT.