xAI's Imagine 2.0, built for images you can ship: a wand-style edit changes the region you point at and nothing else, up to 5 reference images combine in a single generation, and smart resize reframes one image across 5 ratios. It ranks second worldwide in both image editing and text-to-image arenas. On Oddbeam it runs at 30 credits per image, and every text-to-image request returns 2 images.
Imagine 2.0 treats editing as a first-class capability, so the workflow starts from what you need to ship:
Write a prompt up to 4,000 characters naming style, lighting, and exact text, or drop up to 5 reference images — subjects combine, styles transfer, and scenes compose without manual compositing.
Every text-to-image request returns 2 images at 30 credits, and image-to-image edits return 1 refined take — compare variations immediately instead of re-rolling one at a time.
Say the change — recolor the jacket, swap the headline, remove the bystander — and the edit applies to the named region while the rest of the frame stays untouched.
Pick 1:1, 2:3, 3:2, 16:9, or 9:16 and the model fills the new frame — one master image becomes an Instagram post, a YouTube thumbnail, and a TikTok vertical without re-shooting.
What Imagine 2.0 brings to the Oddbeam workbench:
Magic-wand style editing changes the region you describe and preserves everything else — background removal, object swaps, and color changes export clean, ready to drop into other work.
Up to 5 input images in a single generation — combine products, transfer styles, and place characters into new scenes without manual compositing in a separate editor.
2.0 plans layout and typography the way a designer would, so dense multi-part visuals hold together and small text stays sharp — posters, labels, and thumbnails come out usable, not almost-usable.
It preserves what you put in across edits and re-generations — a character, her locations, and her props hold one look from image to image, which is how world-building sets stay coherent.
1:1, 2:3, 3:2, 16:9, and 9:16 cover feed posts, prints, YouTube thumbnails, and vertical short-video covers — resize the master instead of re-prompting per channel.
Workflows that end with a deliverable, not a screenshot.
Five product angles into one composed scene, then smart resize for the feed and the story cover. One master image, five placements, 30 credits a run.
E-commerce Creative
DTC Brand
The two-images-per-run thing changed our review loop — we compare takes in the first pass, and the pointed edits mean revision notes stop being re-shoots.
Social Media Manager
Agency Team
Small text stays legible and the layout holds. I point at the headline, change one word, and export — the background never shifts.
Thumbnail Designer
YouTube Studio
Five product angles into one composed scene, then smart resize for the feed and the story cover. One master image, five placements, 30 credits a run.
E-commerce Creative
DTC Brand
The two-images-per-run thing changed our review loop — we compare takes in the first pass, and the pointed edits mean revision notes stop being re-shoots.
Social Media Manager
Agency Team
Small text stays legible and the layout holds. I point at the headline, change one word, and export — the background never shifts.
Thumbnail Designer
YouTube Studio
Cancel anytime • Unused credits roll over
Perfect for getting started
Great for power users and teams
First choice for professionals
Five references in, two takes out, one master resized for every channel — Grok Imagine Image from 30 credits.