Budget: 120 EUR Deadline: 3 days
Hello Yuri,
I will help you set up a workflow in ComfyUI for high-quality pose extraction from photographs and subsequent image generation based on these poses. I am very familiar with the specifics of running ComfyUI on the Apple Silicon architecture.
Implementation Plan
1. Environment Analysis and Setup:
I will check your current ComfyUI installation on your MacBook M4 and install the necessary custom nodes, primarily ComfyUI-ControlNet and ComfyUI-Impact-Pack.
2. Development of the Core Workflow:
I will create two interconnected workflows:
Workflow #1: Pose Extraction
Tool: The OpenPoseFullPreprocessor node from the ControlNet pack.
Logic: I will configure the pipeline to take your source image and output a clean, detailed pose map (pose keypoints). For complex poses, this may require fine-tuning the preprocessor's parameters.
Workflow #2: Image Generation Based on Pose
Tool: The ControlNetApply node to inject the pose into the generation process.
Integration: I will set up the connection so the output from Workflow #1 is fed into the ControlNet in Workflow #2.
Balancing: I will find the optimal balance between the influence of the pose (control_weight) and the creative freedom of the text prompt. This ensures the pose is preserved accurately while the generated image remains natural-looking.
3. Testing and Optimization:
I will test the workflow on a series of photos with both standard and complex poses.
I will ensure that the pose maps are generated with high quality, without losing key points.
I will optimize the settings for maximum performance on the M4 chip to minimize generation time.
