AI studio
Image, video and voice
Token billing

Glovpix

Sector: Creative AI
Delivered as: Web app
Languages: EN

“One bar. Every tool.”

(The project)

An AI content studio where clients generate and edit images, videos and voice clips, swap faces and backgrounds or animate photos, from one studio interface with token-based billing and a back office that controls tools and pricing.

(Context)

Every model, one studio.

Glovpix gathers text-to-image, image-to-image, text and image to video, video editing, lip sync, virtual try-on, upscaling, background tools and text-to-speech in a single studio, with a prompt assistant that suggests ready-made prompts.

A Next.js front end talks only to a Laravel API, which handles accounts, token billing, pricing and every call to the inference provider. The browser never holds a credential and never talks to the models directly.

(Key figures)

creative tools: 18
media: image, video, voice: 3
text-to-image models: 7
shared generation pipeline: 1

The decisions that shaped it

The browser never sees a credential
Server actions proxy every request to the API with a token kept in an httpOnly cookie, and forward the real client IP so rate limits stay per person.
Secure by design
One pipeline for every tool
Each tool follows the same path: reserve tokens, queue the job with a webhook, then settle it exactly once, whichever of the webhook or the client’s polling arrives first.
Consistent billing
Browse freely, sign in to create
Visitors can open every tool. Sign-in is only requested at the moment they generate, upload or save.
Low friction
Tools and prices run from the back office
Admins enable tools and engines, set image limits and edit pricing tables without touching the studio code.
Operational control

(How it works)

The generation pipeline

Reserve
The request reaches the API, which reserves tokens against the client’s balance before anything billable happens.
Queue
Files are uploaded and the job is queued on the model provider with a webhook address.
Settle once
The client polls every few seconds while the webhook may fire on its own; the first to arrive settles the job, atomically.
Save
The result lands in the client’s creations, and tokens are finalised, or refunded on failure.