The service turns a snapshot into a stylized portrait: realism, anime, cyberpunk. It also removes noise, replaces backgrounds, and builds collages. Photos are taken in a Windows app at the client's location, and generation runs on our Stable Diffusion server. The photos never go to third-party clouds.

Challenge

Receive photos from the app at the shooting location and process them with ready-made templates on our own server, without passing the photos to third-party AI services.

What we did

  1. Our own Stable Diffusion

    We deployed the model on our own server and installed additional SD modules.

  2. Processing API

    The server receives a photo from the app, applies a template, and returns the result.

  3. Windows app at the client's location

    Written in C++ Builder (VCL): shooting, sending to the server, and a template editor.

  4. Templates

    Ready-made presets for styles, backgrounds, and effects.

  5. Collages and retouching

    Combining shots, noise reduction, and sharpening.

Screenshot of AI photo processing on our own Stable Diffusion (#2)
AI photo processing on our own Stable Diffusion: screen 2
Screenshot of AI photo processing on our own Stable Diffusion (#3)
AI photo processing on our own Stable Diffusion: screen 3

Stack

Vue.js, LoopBack (Node.js), PostgreSQL, Stable Diffusion, Docker, C++ Builder.