Image editing workflow
Z Image Turbo画像編集：生成、リミックス、Z-Image-Editの使い分け
text-to-image、image-to-image、指示ベース編集を混同しないための実用ガイド。
2026年7月更新 · チュートリアル + FAQ
チュートリアル
編集パイプラインでZ Image Turboを使う方法
The cleanest workflow separates creative generation from targeted repair. That keeps prompts short and makes each revision easier to judge.
Generate a clean base image
Use Z Image Turbo for fast base concepts. Keep the subject, composition, and lighting simple so the image is easier to edit later.
Remix with image-to-image when structure matters
If you already have a reference composition, use image-to-image and describe what should change while preserving the main pose, product, or layout.
Use targeted editing for local changes
For object removal, background replacement, text swaps, or small product fixes, move to Z-Image-Edit or a masked edit workflow so only the selected region changes.
例
画像編集向けpromptパターン
These prompts separate what should change from what must remain untouched.
Preserve the product
Keep the bottle shape, label, and camera angle unchanged. Replace only the background with a warm bathroom shelf scene, soft daylight, realistic shadows.
Use this pattern when product identity matters more than creative freedom.
Clean object removal
Remove the small coffee cup from the table and reconstruct the wooden surface naturally. Keep all other objects, lighting, and perspective unchanged.
The prompt names the object, fill behavior, and preservation constraints.
Edit-ready base generation
Portrait of a fashion model against a simple gray studio background, clean edges, neutral expression, no busy props, realistic photography, easy to mask
Simple backgrounds make later mask and background edits cleaner.
チェックリスト
編集前チェックリスト
Decide whether the task is new generation, image-to-image, or a masked local edit.
State what must remain unchanged: face, product, text, layout, pose, or lighting.
Use one edit request per pass for reliable comparisons.
Save intermediate outputs so you can roll back when an edit changes too much.
FAQ
Z Image Turbo画像編集FAQ
Can Z Image Turbo edit existing images directly?
Z Image Turbo is primarily used for fast generation. For existing-image edits, use image-to-image, Z-Image-Edit, or a masked editing workflow depending on the precision required.
When should I use Z-Image-Edit?
Use Z-Image-Edit when a specific region needs to change while the rest of the image should stay stable, such as background replacement, object removal, or text edits.
How do I preserve a face or product?
Say exactly what should remain unchanged and keep the edit local. If precision matters, use a reference image or mask rather than a broad text-to-image prompt.
Why did the edit change too much?
The prompt may be too broad, the mask may be too large, or the workflow may be using generation rather than local editing. Narrow the edit and add preservation instructions.
What is the best workflow for ecommerce images?
Generate clean base variants with Turbo, pick the best product framing, then use targeted editing to adjust background, shadows, labels, or props.
Can I use ComfyUI for editing workflows?
Yes. ComfyUI is useful when you need visible node control, saved workflows, and reproducible settings across generation and editing passes.
参考
参考資料と関連ページ
- Z-Image-Edit guide on zimageturbo.ai
Companion page for instruction-based image editing workflows.
- Tongyi-MAI Z-Image-Turbo model card
Reference for Z Image Turbo generation behavior.
生成
編集しやすい元画像を作る
Start with a simple, mask-friendly Z Image Turbo prompt, then move into image-to-image or targeted editing for controlled revisions.