Working deterministic core · Free alpha · Open Studio

A working deterministic core
for cinematic AI media.

Same prompt. Same seed. Same model. Same frame — byte-for-byte. Try the demo right now in your browser. Open Studio loads in a guided tour, walks you through the workspace, and proves the core is real with a one-click determinism check.

Public alpha is free. Render runs on the GPU you control — grab the BYOC starter pack.

What we are building

Where the product stands today.

Each capability below is labelled by status — working today, or in active development. Full detail on the about page.

Deterministic core · working

Same prompt · same seed · same checkpoint, on the same machine and a pinned environment → the same frame, byte for byte. Measured across our SDXL, AnimateDiff and Wan render pipelines. Cross-machine reproducibility is not yet validated.

Cinematic 1080p clips · working

1080p clips at common frame rates and aspect ratios with selectable camera looks — working today. 4K and 8K are the next resolution targets.

Character continuity · working

A face, hair and wardrobe stay stable across cuts and scenes. Character continuity holds in the current pipeline.

Style continuity · working

One style spec governs a whole series so it looks consistent end to end. Style continuity is working.

Provenance manifest · in development

We are designing a C2PA-aligned content-credential manifest for each render, so output can be checked for tampering. Not yet shipped, and not a statement of regulatory compliance. Read more.

Watermark recipe · in development

A post-render watermark recipe (ffmpeg-based) is planned for the BYOC starter pack, with an optional invisible per-recipient mark for leak tracing.

Mission

Lower-carbon by architecture.

Cloud-render incumbents burn datacenter electricity at hyperscale. We render on the GPU you already own, using a deterministic core that avoids wasteful re-rolls. Our own order-of-magnitude estimate points to materially lower energy per finished minute than equivalent cloud workflows.

Order of magnitude only — read how we calculate this on the About page. We do not claim carbon offsets, certifications, or net-zero status. We claim a more efficient architecture.

Studio interface preview — 3-pane workflow with scenes, prompt editor, and deterministic output panel

Studio · 3-pane workflow with deterministic re-render proof

PRE-ALPHA
FREE ALPHA
1080p · cinematic reel
neyralfoundry.com

This is the Studio interface. Click below to open the demo — same prompt + same seed produces a byte-identical render every time. Try the determinism check yourself.

Open Studio — demo mode

Studio design mockups

Where the Studio interface is heading.

The three mockups below show the design direction for Studio — reusable assets, episode-scale orchestration, and an append-only change log. They illustrate the planned interface; these views are in development, not finished features.

Studio Asset Library — character archetypes, styles, cameras, lighting in a grid
Asset library 13 archetypes · 6 styles · 10 cameras · 8 lighting setups. Reusable across the entire series. One reference sheet drives consistent face / wardrobe / proportions across every render.
Studio Episode Timeline — series, episode, scene, shot hierarchy with status indicators
Episode timeline Series → episode → scene → shot. One project file holds 247 shots across 8 episodes. Same style spec, same character library, same camera config — episode 1 looks like itself in episode 8.
Studio Audit Log — append-only provenance event ledger with on-chain verification badges
Provenance audit log Planned: an append-only record of every change — render complete, prompt edit, asset added, bundle exported, scene created. Signing and verification are in development.
How it works

You design. Your hardware renders.

Neyralfoundry is a design surface, not a compute platform. You design scenes, prompts, and pipelines in our environment. You export to a runnable bundle. You run that bundle on whatever GPU you control — any CUDA-capable Nvidia GPU at home, a RunPod pod, an AWS p4d.24xlarge, or your studio's render farm. We never see your inputs or your outputs.

Why this shape

  • Your data never leaves your perimeter, so a studio security team has far less to review — the architecture is built to be quick to approve.
  • Our compute bill is zero, so we can keep this free during alpha without burning capital.
  • You keep full control of model choice, version, and any fine-tuning you bring.
  • We focus our engineering on the design surface and the orchestration spec, not on running someone else's GPU fleet.

What is in the early-access alpha

  • Scene treatment surface (storyboard + prompt builder + camera/lighting/style configs).
  • Export to runnable bundle compatible with ComfyUI, AUTOMATIC1111, diffusers, and similar local pipelines.
  • Output watermark embed helper (post-render ffmpeg recipe).
  • Quickstart starter pack for a single 60-second reel.

What is not yet shipped

  • Cloud rendering (deliberate — will not ship before we have funding to operate it safely).
  • Multi-user collaboration (planned for beta).
  • Asset library + reusable character system (planned for beta).
  • SOC 2, full compliance suite (planned post-beta).
Partnerships & press

Other ways to reach us.

For partnership, press, or general inquiries, use the contact form. We read everything and reply when there is a fit.

Ready to try it?

Join the waitlist or grab the starter pack now and run on your own hardware.

Join waitlist