How to use LLaDA-Image
How to Use LLaDA-Image Online
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.
This site is an independent third-party browser workspace for LLaDA-Image. It is not affiliated with inclusionAI. Official code and weights remain on GitHub, Hugging Face, and ModelScope.
A practical first session
Open the workspace, choose LLaDA-Image (Base) or LLaDA-Image Turbo, then either write a prompt or upload a reference to edit. This page walks through that flow so you can get a usable result without local GPU setup.
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
Keep prompts concrete, set expectations for Base vs Turbo, and verify model claims against the official GitHub and paper when you share results.
- 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.
Related pages
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.