-
Flux Kontext + Ollama in ComfyUI
Read This Post: Flux Kontext + Ollama in ComfyUII designed a lightweight LLM-powered workflow to improve consistency and quality in image prompting. The process begins with Microsoft’s Florence 2 vision model, which analyzes and describes the content of the input image. This description is then combined with user input using Ollama to create a richer, more context-aware prompt. Finally, the refined prompt is…
-
Simple ComfyUI Web App Using ViewComfy and Ngrok.
Read This Post: Simple ComfyUI Web App Using ViewComfy and Ngrok.I’ve created a simple locally run ComfyUI web app using ViewComfy and Ngrok.
-
Replicate Anything with ControlNet, IPAdapter, and Florence2 CLIP Encoder in ComfyUI.
Read This Post: Replicate Anything with ControlNet, IPAdapter, and Florence2 CLIP Encoder in ComfyUI.In this project, I used a custom SDXL checkpoint, ControlNet with DepthAnythingV2, IPAdapter with style transfer, and the Florence2 open-source vision model to replicate the original photo. By integrating these three methods, the final image aligns with the original in terms of style, depth map, and descriptive prompt. How It Works𝐂𝐨𝐧𝐭𝐫𝐨𝐥𝐍𝐞𝐭 𝐰𝐢𝐭𝐡 𝐃𝐞𝐩𝐭𝐡𝐀𝐧𝐲𝐭𝐡𝐢𝐧𝐠: This tool…
-
Living Room Study with Flux-Dev (NF4) in ComfyUI.
•
Read This Post: Living Room Study with Flux-Dev (NF4) in ComfyUI. -
Smart Flux Inpainting Workflow With Automated Visual Recognition and Masking in ComfyUI.
•
Read This Post: Smart Flux Inpainting Workflow With Automated Visual Recognition and Masking in ComfyUI.In this workflow, I used Florence and Segment Anything vision model to automatically recognise elements in the image and create masking layer based on the text prompt. I applied SEGS Detailer and Flux to complete the inpainting.
-
Virtual Reality Flux Workflow With ComfyUI + Kuula.
Read This Post: Virtual Reality Flux Workflow With ComfyUI + Kuula.Link to VR tour in Kuula: https://kuula.co/share/collection/7ZlgS
-
LLM Generated HTML Webpages with OpenManus.
•
Read This Post: LLM Generated HTML Webpages with OpenManus.I used OpenManus, an open-source, general-purpose autonomous AI project, alongside the Anthropic Claude 2.5 LLM model to generate responsive HTML webpages from simple prompts.
-
Locally Run LLM With Browser Use.
•
Read This Post: Locally Run LLM With Browser Use.Integrating the locally run Qwen2.5:7b language model with Browser-Use through Ollama enables efficient automation of browser tasks while maintaining data privacy. While the integration offers numerous advantages, it’s important to note that smaller models may occasionally produce incorrect output structures, leading to parsing errors. To maximize Browser-Use’s potential, I recommend using ChatGPT-4o API support.
-
Browser Use With ChatGPT-4o API.
•
Read This Post: Browser Use With ChatGPT-4o API.With Browser Use running locally in Docker and powered by the ChatGPT-4o API, I’m exploring how AI can automate everyday tasks using computer vision and reasoning.
