Z-Image + Qwen Edit workflow for consistent-face AI selfies in ComfyUI
tomxp411 · reddit · 2026-09-27
The author shares a complete ComfyUI workflow combining Z-Image-Turbo with Qwen Image Edit 2511 to generate selfies with consistent facial features, with the workflow file on Google Drive.
- Z-Image was chosen for best general-purpose output, but it can't use reference images for face consistency, so Qwen Edit 2511 (which shares components with Z-Image Turbo) handles the face swap.
- Steps: open a Z-Image-Turbo text-to-image workflow, right-click-copy the Text to Image node into the Qwen Image Edit workflow (right-click paste, not Ctrl+V); connect text-to-image IMAGE output to image1 and the reference photo to image2 on Image Edit.
- Prompt template: Swap the head from <image2> onto the person in <image1>. Preserve the lighting and shading from <image1>.
- Tip: add a Preview Image node to iterate on base images cheaply, then run the final Save Image node for the face swap.
The long-term goal is a local AI ecosystem combining 1:1 chat, group chats and image generation on one system.
More from Multimodal
- Wanted to film a software quality series, made AI songs instead — EricBuess · 2026-09-27
- Maker Created 100 Videos With Claude Opus 5.5, Open-Sources 39 Style Prompt Library — dotey · 2026-09-27
- Leak: Meta Muse to announce four more partnerships by end of October, first teased as Runway — ChrisGPT · 2026-09-27
- Ming Image impresses Reddit with long-text rendering, called a top open-source image model — Philosopher_Jazzlike · 2026-09-27
- Blogger shares Claude Opus 5.5 video generation tests with 54 prompts — vista8 · 2026-09-27
- Claude Opus 5.5 rigs and animates a 3D print in Blender from a single prompt — Worth-Sir-8756 · 2026-09-27