Designing an AI image + video workspace in the browser: credit costs, task history, and fewer tabs

Most AI image tools are single-purpose: one page to generate, another to edit, another to make a video, and a fourth to upscale the result. When I started working on Uyo, the constraint I cared about was simpler than any model choice: keep the whole loop inside one browser workspace, and make the cost of each action obvious before the user clicks.

What the workspace actually does

• Text to image - describe a scene and generate results.
• Image to image - start from a reference photo instead of a prompt alone.
• Upscaling and background removal - cleanup steps that used to mean opening a second tool.
• Text to video and image to video - turn a prompt or a single still into motion.

It runs in the browser, so there is no installer and no platform checklist. The interface is available in English and Chinese.

Show the price before the request

The part I keep coming back to is credit visibility. Image and video generation share one balance, and the cost varies by model and output settings. That makes "generate" a small financial decision, not just a creative one, so the workspace displays an estimated credit cost before generation instead of after.

If you have ever queued a batch of 20 renders and only then noticed the balance, you know why this matters. It also removes a whole category of support questions.

Keep the output, not just the prompt

Every generation lands in My Creations, along with the task itself. In practice that means the history is the product: you come back to a result you made last week, not to a prompt you half remember. Keeping tasks and results together also makes the workspace usable on a laptop where you close the tab constantly.

Pricing, stated plainly

Uyo is a paid product. Credits are sold as one-time packs:

• 400 credits for $9, valid 30 days
• 2,500 credits for $49, valid 90 days
• 6,000 credits for $99, valid 365 days

Public tools, examples and guides can be browsed without an account. Generation requires an account and sufficient credits. I would rather write that in the first paragraph of a post than bury it behind a modal.

What I would change next

Three things are still rough:

1. Cost estimation for mixed workflows. A single image is easy to estimate; an upscale plus a background removal plus a video pass is not. Showing a running total per project would help.
2. Better reference management. Right now a reference photo is per task. A small library of recurring references would save retyping URLs.
3. More explicit model notes. Different models behave differently on the same prompt, and the UI should say so rather than implying one universal result.

None of these are exotic. They are the kind of thing you only notice after using your own tool for a few weeks, which is exactly why shipping early mattered.

Try it

If you want to see the workspace, it is at uyo.io. If you build something similar, I would be interested in how you handle credit estimation in the UI - that seems to be the part every tool gets slightly wrong.

Disclosure: I work on Uyo.