Run Replicate

Start by pasting your Replicate API key, then pin the few image models you actually use so they’re always one click away. From there, you can run quick prompt tests, adjust parameters, and re-run variations without losing your place or hopping across multiple pages. When a result is close, switch to a different model for the next step—such as a refiner or an upscaler—and keep moving through the same session as you iterate.

For repeatable work, set up a simple sequence that matches how you generate in real life. A common flow is to create a base image, pass it through a refinement step, then upscale for final delivery. Once the chain is saved, you can run it again with new prompts or small tweaks to settings and compare outcomes across versions. This is useful when you’re producing a batch of assets in a consistent style and need the same processing steps every time.

As you generate, keep outputs inside the app so you can review what worked, spot differences between runs, and return to earlier versions without digging through folders. That makes it easier to pick finalists for clients, maintain a clean history during long prompt sessions, and avoid overwriting good results. For extended work sessions, switch the theme to stay comfortable while tuning prompts and model settings.