Four workflows, one chat input. Each clip below is the actual editor, sped up — real prompts, real outputs, just compressed past the parts where you'd be waiting on the model. Pick the one that fits the next thing you have to ship.
1. Create an Image From Text
Write a sentence, get an image. The model reads the prompt and picks an aspect ratio it thinks fits, or you pin 2:3, 1:1, or 3:2 from the tool tray yourself. Generations are cheap enough that you can run the same prompt five times in a row, changing two words each pass, until the composition actually lands.
Reach for it when: a stock photo would look stock. Hero images, blog covers, social posts, character concepts, mood frames before a shoot.
2. Edit an Existing Image
Drop in a photo, describe the change. The subject — person, product, character — stays recognisable. Everything around it gets rewritten: clothing, background, lighting, pose, time of day, whatever you wrote down. No masks, no layer stack, nothing to learn beyond the sentence.
Reach for it when: you need to turn one shoot into a dozen variants, keep a character on-model across a series, or rebuild a thumbnail without losing the face that's already earning clicks.
3. Turn a Photo Into a Video
Hand the model a still and it invents the next few seconds. Your photo is frame zero; everything after that is generated motion — a head turns, eyes blink, a product rotates, wind catches the trees in the background. The output is a normal video file you can trim, caption, or cut into something longer.
Reach for it when: a static asset works but feels flat. An ad creative that should breathe, a landing hero that should move, a portrait that should glance at the camera.
4. First → Last Frame Video
Select two images on the canvas. The model treats one as the opening frame and the other as the closing frame, then invents the motion that connects them. You write the bookends; the model fills the middle.
Reach for it when: you need a transition that looks intentional. Before-and-after reveals, character transformations, day-to-night shifts, product evolution across a campaign.
All four live behind the same chat. You can move between them mid-session — no mode picker, no separate tool, no menu to relearn. Just describe what you want next.