SDXL + Z-Image Turbo: a 2-step hybrid workflow in ComfyUI with CMK Flow
Away-Sheepherder-578 · reddit · 2026-10-26
- A hybrid workflow in CMK Flow: SDXL handles composition and initial generation, then Z-Image Turbo refines it in just 2 sampling steps at 0.15 denoise — no face restoration or detailer needed.
- A demo video walks through the ComfyUI setup, generation process, and close-up comparison of first-pass vs. refined images.
- CMK Flow is a free, open-source collection of modular ComfyUI custom nodes, guided workflows, and a Flow Browser, available on GitHub (v2.6.1) and Comfy Registry.
More from Multimodal
- Imagination Studio: rebuilt local open-source platform for video, audio, music and image generation — Scobleizer · 2026-10-26
- Beeble SwitchX 2 swaps video backgrounds and relights footage so well this creator is ditching the podcast studio — venturetwins · 2026-10-26
- Cinematic AI music video for original song 'Gaslight' — Hellish_NDE · 2026-10-26
- Smallest Pulse STT transcribes Eminem's fastest verse: 296 words at 1.4% WER — auto_grad_ · 2026-10-26
- UniForge3D open-sourced: joint 3D reconstruction and generation from sparse multi-view images — victormustar · 2026-10-26
- AI-generated MV 'Wings of Miracles' for virtual character Kei Kirishima — No-Past-7449 · 2026-10-26