ComfyUI workflow
Workflow ComfyUI Z Image Turbo : installation, réglages et dépannage
Utilisez le tutoriel officiel ComfyUI comme base, puis ajustez prompt, seed et résolution.
Mis à jour en juillet 2026 · guide + FAQ
Guide
Un workflow ComfyUI stable pour Z Image Turbo
ComfyUI is powerful because every node is visible. That also means small changes can hide the source of a problem, so start with a minimal baseline.
Install from the official workflow path
Follow the ComfyUI Z-Image Turbo tutorial for current model and workflow requirements. Confirm the model files are in the expected folders before changing the graph.
Run a baseline generation
Load the workflow, enter a short prompt, keep a fixed seed, and generate at a standard square resolution. Save the output and the settings before tuning.
Tune prompts before tuning nodes
Most quality changes should start in the prompt. Only adjust steps, resolution, or advanced node settings after you know the prompt is clear.
Exemples
Prompts pour tester le graphe ComfyUI
These prompts exercise common weak points: materials, text, composition, and lighting.
Material detail
Close-up product photo of a brushed aluminum speaker, softbox reflection, dark studio background, crisp edges, realistic texture, 1024px
Checks whether the workflow preserves surface texture and clean edges.
Poster typography
Modern launch poster, white background, large readable title "TURBO MODE", small orange geometric accents, minimal Swiss typography
Use short text to test typography behavior before adding more layout complexity.
Seed comparison
Cozy reading corner, warm lamp, wooden chair, rain outside the window, cinematic interior photography, balanced composition
Run this at several seeds while keeping every other setting unchanged.
Checklist
Checklist de dépannage ComfyUI
Confirm model files, custom nodes, and workflow JSON match the tutorial version.
Restart ComfyUI after installing nodes or moving model files.
Use a fixed seed while comparing prompts or settings.
Lower resolution first if memory errors appear, then scale up after the graph is stable.
FAQ
FAQ Z Image Turbo ComfyUI
Where should I start?
Start with the official ComfyUI Z-Image Turbo tutorial and its current workflow. Avoid mixing older graphs until the baseline runs.
Can I use the same prompts online and in ComfyUI?
Yes. Prompt structure transfers well, but local settings such as seed, resolution, and workflow nodes will affect the final image.
What should I change first when quality is poor?
Change the prompt first. If the prompt is clear and quality is still poor, adjust resolution, seed, and workflow settings one by one.
Why does ComfyUI run out of memory?
Resolution, batch size, model precision, and GPU VRAM all matter. Reduce resolution or batch size first, then retry.
Do I need ComfyUI to use Z Image Turbo?
No. ComfyUI is useful for local node workflows, but the Image Creator AI workspace is faster for web-based prompt testing.
How do I keep results reproducible?
Save the workflow JSON, prompt, seed, model version, resolution, and any node settings used for the final output.
Sources
Sources et lectures associées
- ComfyUI official Z-Image Turbo tutorial
Official ComfyUI tutorial for installing and running a Z-Image Turbo workflow.
- Tongyi-MAI Z-Image-Turbo model card
Model reference for weights, architecture context, and usage notes.
Cluster
Continuer le dossier Z Image Turbo
Z Image Turbo prompts
Write cleaner prompts before changing the workflow graph.
Lire ensuiteZ Image Turbo image editing
Decide when to use Turbo, image-to-image, or Z-Image-Edit.
Lire ensuiteZ Image Turbo vs Nano Banana Lite
Compare local/open workflow fit against a lightweight Gemini path.
Lire ensuiteGénérer
Prototyper le prompt en ligne
Use the hosted workspace to check prompt direction quickly, then move winning prompts into your ComfyUI graph for deeper control.