NeuroCognica
The product
You describe a thing. A language model on your machine turns that into a scene plan. The plan is validated, built as real geometry in Blender, rendered, and sealed. Every step happens on your own hardware.
What happens when you press Create
| Step | What it does |
|---|---|
| Sentinel review | Your request is checked against the published prohibited-use policy before anything is generated. Cleared runs say so; refused ones say why and give you an appeal route. |
| Imagining | A local language model reads your sentence and proposes a scene — what the object is, what it is made of, how it should sit and be lit. |
| Planning | That proposal is compiled into a typed scene plan and put through validation gates. A goblet cannot come back as a torus; glass stays glass. |
| Materials | Textures are synthesised locally to match the material intent rather than pulled from a stock library. |
| Painting | Where the presentation calls for artwork — the museum wall, the canvas on an easel — an image model paints it locally. This is the longest step and the progress bar says so honestly. |
| Building | Blender constructs real procedural geometry from the plan and renders it. |
| Sealing | Prompt, plan, model, timings and every policy decision are written into a hash-chained ledger, and the bundle is assembled. |
The progress you see is not a guess. The engine reports its own position as structured events, so a step that has ended is never still displayed as running, and a step that was skipped is reported as skipped rather than as a green tick.
Presentation modes
Museum Wall
Your work framed on a gallery wall under track lighting, with an information placard naming the title, medium, inspiration and accession. The default presentation.
Showroom
A controlled white studio with a lit pedestal — the object alone, evenly lit, from locked camera angles. Best when the thing itself is the subject.
Canvas on an easel
A tripod easel holding a framed canvas, painted by a live local render of your scene, lit so it reads crisply against the studio exposure.
Named recipes
Deterministic constructions — a crystal spire, a floating jawbone — selected by the display control or by a word in your prompt.
The output is yours and it is editable
Every render produces three things: a PNG, a real .blend with
procedural geometry, and a structured manifest. The .blend opens in
Blender 4.5 with no dependency on this software — move the objects, change
the materials, relight it, animate it. It keeps working if you never run
ChronoSophia again.
We claim no ownership of anything you make. No licence back to us, no portfolio rights, no royalties. Your work, including the ledger, is yours — see section 4 of the licence.
Local means local
There is no cloud inference in the creative path, no account, and no phone-home. The only network activity is the optional download of tools and models during setup, which you consent to item by item.
Practical consequence: it works offline, it cannot be discontinued out from under you, and nothing you make is training data for anybody.
Provenance, and what it does and does not prove
Artifacts are sealed into an append-only, BLAKE3 hash-chained ledger on your disk. Modify a past entry and the chain breaks detectably. Images also carry an embedded machine-readable declaration that they were AI-generated.
Two honest limits: that declaration is metadata and can be stripped by re-encoding the image, and we make no claim of C2PA conformance. It is an honest label, not an unforgeable one. The ledger proves what system generated this artifact — not the intent behind the prompt, and not that the result is good.
What it is not
It makes staged, single-subject scenes well. It is not a universal world generator and not a replacement for a modelling suite. If you need arbitrary complex environments authored to specification, this is the wrong tool and we would rather say so now.