Building OmniEnhancer: From a Blurry Photo to a Publish-Ready Visual

hua Huang — Sun, 09 Aug 2026 10:32:57 +0000

Image enhancement sounds simple until you try to turn it into a repeatable workflow.

A typical creator or ecommerce seller may start with a compressed product photo, a soft portrait, or an old image. Improving it often means jumping between an upscaler, a background remover, a photo editor, and a generation tool. Every switch adds another upload, another set of controls, and another place to lose track of the best result.

That fragmentation is why I built OmniEnhancer: a browser-based workspace for improving image quality and continuing directly into the next creative task.

The workflow I wanted

The core flow is intentionally straightforward:

Upload an image in JPG, PNG, or WebP format.
Enhance the photo to improve clarity, sharpness, color, and visible detail.
Upscale the result when a larger export is needed.
Keep editing with tools such as background removal, object cleanup, face enhancement, or old-photo restoration.
Reuse the improved asset for product scenes, ads, thumbnails, social posts, or image-to-video experiments.

Instead of treating enhancement as the final step, OmniEnhancer treats it as the start of a broader visual workflow.

You can try the focused AI Image Enhancer or go directly to the Image Upscaler.

Why one workspace matters

For ecommerce, a single source image may need to become:

a cleaner marketplace listing;
a square social post;
a transparent-background product cutout;
an ad creative with a new scene;
a sharper thumbnail;
or a short product video.

Creators face the same problem with portraits, screenshots, thumbnails, and archive photos. The hard part is rarely one isolated edit. The hard part is moving from a weak source asset to several useful outputs without rebuilding the workflow each time.

OmniEnhancer keeps those related tasks together, so the enhanced image can become the input for the next operation.

A few engineering decisions behind it

The web app is built with TanStack Start, React 19, and TypeScript. Client-side data access uses TanStack Query over typed API helpers, while server routes call a provider abstraction for different AI models.

A few implementation details became especially important:

1. Normalize model differences

Image and video providers expose different parameters, response shapes, and job states. The UI should not force users to understand every provider-specific detail. A shared task contract presents consistent states such as queued, processing, completed, and failed.

2. Make cost visible

A quick preview and a high-resolution model do not have the same compute cost. OmniEnhancer uses a credit-aware task flow so users can see the tradeoff before starting a premium operation. Failed jobs are handled through the same tracked lifecycle instead of becoming invisible billing errors.

3. Treat privacy as a product control

Some users want cloud history; others are working with unpublished product shots or personal photos. Private Mode is designed to avoid saving the upload, prompt, settings, or result to OmniEnhancer's cloud history while clearly explaining that the processing provider must receive the file to run the selected model.

4. Keep focused tools and an advanced studio

A dedicated enhancer page is faster for a single job. A studio is better when the user wants model selection, reference assets, generation, editing, or video. Supporting both avoids turning every simple cleanup task into a complicated dashboard session.

What I learned

The most useful lesson was that "AI image enhancement" is not one feature. It is a chain of decisions:

What is actually wrong with the source image?
Should the system sharpen, denoise, restore, or upscale?
Does the user need a fast preview or a high-quality export?
What will they create after the image is improved?
How much history should be stored?

Designing around that chain produces a more useful product than exposing a long list of disconnected models.

Try it and share feedback

If you work with product photos, creator assets, portraits, or old images, you can explore OmniEnhancer.

I would especially like feedback on two questions:

Which enhancement task causes the most friction in your current workflow?
After improving an image, what is the next tool you usually open?

Those answers will help me decide which workflows deserve the most attention next.

DEV Community: hua Huang