Comment utiliser LLaDA-Image en ligne
Comment utiliser LLaDA-Image en ligne
A practical walkthrough for first-time users: open the workspace, pick Base or Turbo, generate or edit, then verify claims against official GitHub and Hugging Face.
Ce site est un espace de travail tiers indépendant pour LLaDA-Image. Il n’est pas affilié à inclusionAI. Le code et les poids officiels restent sur GitHub, Hugging Face et ModelScope.
Tutorial intent, not hype
This guide explains how to use LLaDA-Image through this independent site. It is not an official inclusionAI manual. For sampling hyperparameters and training recipes, always cross-check the public repository and paper.

Step-by-step
Use these steps the first time you open the site.
Enter the workspace
From Home or this page, open the generator. Confirm whether you want Base-quality or Turbo-speed labeling in the model switcher.
Generate or upload
For how-to-use basics: start prompt-only. For edits, upload a reference and write one clear instruction.
Save, compare, price
Keep a short prompt log. If you need more volume, open Pricing. Cite official sources when sharing results publicly.
Common mistakes to avoid
Vague prompts, mixing Base/Turbo expectations, and treating this UI as the official demo are the usual traps. Stay explicit about independence and verify numbers from primary sources.
- Do not invent scores or call this the official demo.
- Match aspect ratio and size guidance to the mode you use.
- Use Turbo for drafts, Base when detail matters.
Pages associées
How-to FAQ
Where do I start if I only have a text idea?
Open the generator workspace, write a detailed prompt, and generate with Base first if quality matters.
How do I edit an existing image?
Upload it, describe the change, and keep identity constraints in the instruction. See the editor page for a focused walkthrough.
How do I learn Turbo vs Base?
Read the Turbo page and the Turbo vs Base blog post, then try both on the same prompt.
How do I cite official materials?
Link GitHub inclusionAI/LLaDA-Image, Hugging Face checkpoints, and arXiv:2609.03796.
Practice the workflow
Open the workspace with one prompt and one edit. Then explore prompts and Turbo pages for deeper intent pages.