A studio for people who
ship the work.
Ideovex is a generative studio that puts image, video and music models behind one prompt box, one credit balance and one asset library — with the price of every render printed on the button.
Why we built it
Generative models got extraordinarily good in a very short time. The tooling around them did not keep up. Anyone doing real production work ended up with a browser full of tabs — one lab for photoreal stills, another for typography, a third for motion, a fourth for audio — each with its own account, its own billing, its own idea of what a prompt is, and no shared history.
Worse, almost none of them told you what a render was going to cost until after it had run, and none of them gave the credits back when the render failed.
Ideovex exists to collapse that. One place to compose, one balance to spend, one library to search, and enough visibility into the machinery that you can make an informed decision about which model to reach for and what it will cost you.
What we are not
We do not train foundation models, and we do not pretend to. Our job is the layer above: making a large, fast-moving catalogue of models genuinely usable — comparable, priced honestly, and wrapped in a workflow that survives a real deadline.
Where we are going
The near-term roadmap is about depth rather than breadth: reusable character references that stay consistent across a whole set, guided recipes for the jobs people repeat every week, side-by-side model comparison, and a public API so the studio can sit inside your own pipeline. The changelog is the honest record of what actually ships.
Show the price before the click
Every render carries a visible credit cost, and a failed job refunds itself. Nobody should discover what a tool costs from an invoice.
One studio, not twelve tabs
Image, video and music models from different labs, behind one prompt box, one balance and one asset library. Switching tools should never be part of the creative process.
No black boxes
Queue position, concurrency slots, generation latency and per-model success rates are all visible. If something is slow, you get to see why.
Private until you say otherwise
Nothing you generate is published, promoted or used as an example unless you explicitly share it. Sharing is a decision, not a default.
Under the hood
Three things we actually build.
The interesting engineering in a generative studio is not the model — it is everything that has to be true around it.
The model layer
A single catalogue that spans image, video and music models, normalised so a prompt, a reference image and an aspect ratio mean the same thing everywhere. When a provider degrades, models are flagged and hidden rather than left to fail on you.
The studio
Prompt composer, searchable model gallery with real sample output on every card, reusable presets, batch runs, and result actions that chain — recreate, use as reference, animate.
The plumbing
Atomic credit accounting, a priority render queue, live job status over websockets, share links with access levels, and analytics that tell you where your credits actually went.
A small team, shipping in public.
We are a small, independent team. That has consequences we are happy to be judged on: we ship narrow features that work rather than broad features that half-work, we answer support ourselves, and we write down what changed every time we deploy.
110+
Models in the catalogue
Image · Video · Music
Output types
Automatic
Credit refunds on failure
Every release
Public changelog
See whether it fits how you work.
Free credits on signup, the full model library unlocked, and nothing to cancel if it is not for you.