DreamOmni2 is a browser-based studio for advanced image editing and generation from plain-language text instructions. Upload up to two reference images, describe the result you want, and generate a new image while keeping key details consistent.
Multimodal control with text + references
DreamOmni2 combines prompts with visual examples. Reference images can guide style, camera angle, or character identity, while your instruction specifies what to change. This helps make targeted edits without breaking the original look.
- Use references to set style, viewpoint, or subject identity
- Adjust background, clothing, lighting, or composition via text instructions
- Mix styles while keeping important elements recognizable
Precise edits and complex transformations
The tool is designed for instruction-based editing: you can request specific changes, blend multiple styles, increase similarity to a reference, or intentionally move away from it. Because the model processes the request as a whole, larger transformations tend to stay visually coherent.
Works fully online
Everything runs online with no software installation. Supported formats include PNG, JPG, JPEG, and WEBP up to 10 MB. Each generation uses credits, making it easier to track usage in a workflow.
