Every model.
One canvas.
Type your intent — the auto-router picks the right model for you. Image, video, edits, upscales, all from one credit pool. You can always override.
Featured this month.




gpt-image-2
NEWimageBest for editorial portraits, product photography, and brand work that needs precise typography. Understands camera vocabulary and renders in-image text reliably.




nano-banana-2
NEWimageThe latest flash Nano Banana (Gemini 3.1) — faster than Pro, sharper than the base model. The default for iteration-heavy workflows and image-to-image refinement.




nano-banana-pro
imageStrong at art-directed scenes — controlled composition, consistent style across multi-image series, and brand-aligned palettes. The picker reaches for it when 'editorial layout' or 'consistent set' shows up in the brief.




Seedream 5.0 Lite
imageByteDance Seedream 5.0 Lite — artistic, painterly output with quick turnaround at low cost. Best for illustration, character art, and stylized ideation passes.




Qwen Image 2.0
NEWimageAlibaba's Qwen Image 2.0 renders legible in-image text and complex layouts straight from the prompt — posters, infographics, slides, comics, product labels. Best-in-class at both Chinese and English typography.




Ideogram V3
NEWimageIdeogram V3 is a typography and graphic-design powerhouse — bold posters, brand marks, quote cards, and editorial layouts with crisp, legible in-image text and a strong design sense.




Imagen 4
NEWimageGoogle's flagship Imagen 4 — fast, natural photorealism for portraits, product shots, landscapes, and food, with a clean look and reliable typography.




Recraft V3
NEWimageRecraft V3 is built for design work — clean logos, icon sets, flat illustrations, and vector-style artwork with consistent strokes and brand-ready output.
Image models 8








Video models 10










Don't pick a model.
Describe the work.
The auto-router reads your intent — modality, style, references, output spec — and picks the cheapest model that meets the bar. Override with the model picker any time.
You type: "a luxury watch, dark moody studio, top-down"
Router classifies: photoreal product. Picks gpt-image-2 over Seedream 4.5 since the text-in-image bar matters here.
4 variants in 1.4 seconds. Drag to canvas, mask, expand, re-prompt — repeat with the same or any other model.
Specs at a glance.
| MODEL | TYPE | MAX RES | SPEED | STATUS |
|---|---|---|---|---|
| gpt-image-2 | image | 4096² | ~1.4s | live |
| nano-banana-2 | image | 2048² | ~0.9s | live |
| nano-banana-pro | image | 2048² | ~1.6s | live |
| Seedream 5.0 Lite | image | 1536² | ~1.4s | live |
| Qwen Image 2.0 | image | 2048² | ~2s | live |
| Ideogram V3 | image | 2048² | ~8s | live |
| Imagen 4 | image | 2048² | ~3s | live |
| Recraft V3 | image | 2048² | ~5s | live |
| Veo 3.1 | video | 1080p · 8s | ~90s | live |
| Veo 3.1 Fast | video | 1080p · 8s | ~45s | live |
| Kling v3 | video | 1080p · 10s | ~80s | live |
| Kling v3 Omni | video | 1080p · 10s | ~85s | live |
| Kling O1 | video | 1080p · 10s | ~80s | live |
| Seedance 2.0 | video | 1080p · 12s | ~70s | live |
| Seedance 2.0 Fast | video | 1080p · 8s | ~40s | live |
| Seedance 2.0 Mini | video | 720p · 5s | ~30s | live |
| Hailuo 2.3 | video | 1080p · 6s | ~55s | live |
| Vidu Q3 | video | 1080p · 8s | ~60s | live |