How it works
- Choose a model from the picker. Recommended models are marked; NSFW-capable models carry a label.
- Write a prompt, or let the Prompt Agent write one (Prompt Agent).
- Optionally add reference images: uploads, images from your history, or a saved character. The maximum number of references depends on the model.
- Set aspect ratio or size, resolution (1K, 2K or 4K where supported), quality tier and the number of images (up to 4 per request; Kling Image O1 up to 2).
- Check the credit cost next to the generate button and generate.
Results appear at the top of your history with live progress.
Modes
- Text-to-image. Generate from a prompt alone.
- Image editing / image-to-image. Add one or more reference images and describe the change: replace objects, change backgrounds or outfits, restyle, combine several references into one scene. Most models switch to edit mode automatically when references are present.
- Character references. Pick a saved character and the model receives its reference photos, keeping the same person consistent across generations.
- Model-specific options. Negative prompts, seeds, web search (Nano Banana 2), face-control elements (Kling), quality tiers (Seedream, Wan, GPT Image).
Image to Prompt
Reverse-engineer any image into a reusable prompt. Upload an image or pick one from your history, choose plain text or structured JSON, and fast or pro:
- Fast (text or JSON): quick analysis at a low cost.
- Pro (text or JSON): more detailed analysis at a higher cost.
The result is inserted into the prompt field and can be edited before generating.
History, albums and sharing
- History keeps every result. Search it and open any image to see its prompt and settings.
- Pin favourites so they sort first.
- Albums group images per project or client; add an image to an album from its menu.
- Public or private. Generations are private by default; toggle individual images public.
- Download originals, or reuse an image as a reference for the next generation.
- Delete images you no longer need to free up storage.
Image models
Every model with its modes, reference support, resolutions and aspect ratios. The studio shows the credit cost of a generation before you start it. "Mature" marks models that are NSFW-capable for adult creators.
| Model | Vendor | Modes and references | Resolution | Aspect ratios | Options |
|---|---|---|---|---|---|
| Seedream 5.0 | ByteDance | Text-to-image and editing; up to 6 reference images or a character | 1K, 2K (Pro) | 1:1, 3:4, 9:16, 4:3, 16:9, auto | Lite or Pro quality. Mature. |
| Nano Banana 2 | Text-to-image and editing; up to 6 reference images or a character | 1K, 2K, 4K | auto, 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16 | Web search. Recommended. | |
| Nano Banana Pro | Text-to-image and editing; up to 6 reference images or a character | 1K, 2K, 4K | auto, 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16 | Strong text rendering. Recommended. | |
| Seedream V4.5 | ByteDance | Text-to-image and editing; up to 6 reference images | up to 4K (auto 2K or 4K) | 1:1, 3:4, 9:16, 4:3, 16:9 | Seed. Mature. |
| Influencer Remix | SocialArt | A saved character or up to 6 reference photos; no prompt needed (optional) | 1K, 2K, 4K | auto, 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16 | Finished influencer-style scenes. |
| GPT Image 2 | OpenAI | Text-to-image and editing; up to 6 reference images | up to 2048×2048 | 9:16, 3:4, 1:1, 4:3, 16:9 | Low, medium or high quality; PNG, JPEG or WebP. Recommended. |
| Wan 2.7 | Alibaba | Text-to-image and editing; up to 6 reference images | 1K | 1:1, 3:4, 9:16, 4:3, 16:9 | Standard or Pro quality. Mature. |
| Nano Banana | Text-to-image and editing; up to 6 reference images | 1K | auto, 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16 | Fast everyday generation. | |
| Qwen Image 2 | Qwen | Text-to-image and editing; up to 3 reference images | 1K | 1:1, 3:4, 9:16, 4:3, 16:9 | Pro tier; negative prompt, seed. Mature. |
| Kling Image V3 | Kling | Text-to-image, image-to-image (1 reference); up to 2 face-control elements | 1K, 2K | 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9 | Negative prompt. |
| Grok Imagine | xAI | Text-to-image and editing (1 reference) | 1K | 2:1, 20:9, 19.5:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16, 9:19.5, 9:20, 1:2 | Ultra-wide and tall formats. |
| Kling Image O1 | Kling | Up to 6 references addressed as @Image1, @Image2; up to 2 elements as @Element1, @Element2 | 1K, 2K | auto, 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9 | Up to 2 images per request. |
| Z-Image Turbo | Tongyi-MAI | Text-to-image only | Standard (about 1 MP) or Max (up to 1440 px) | 10 ratios | Steps, guidance and seed; JPG, WebP or PNG. Mature. |
| GPT Image 1.5 | OpenAI | Text-to-image and image-to-image; up to 4 reference images | 1K | 1:1, 2:3, 3:2 | Auto, low, medium or high quality; WebP, PNG or JPEG. |
Which image model should I use?
- Speed and value: Nano Banana 2, Grok Imagine, Z-Image Turbo.
- Top quality and text rendering: Nano Banana Pro, Seedream 5.0 Pro.
- Precise edits and instruction following: GPT Image 2, Kling Image O1.
- Consistent people: Kling Image V3 elements, or any model with a saved character.
- Influencer content without prompting: Influencer Remix.