Upload your CAD, references or script. Agents stage the scenes, check every frame against your source, and fix what the AI invents.
Production-proven in automotive
For FIFA World Cup 2026, Volkswagen ran its first AI social pilot — 29 assets across Instagram, TikTok, Threads, and LinkedIn. In automotive, every published image must be product-perfect; VW's bar is among the highest. SolidMeta's 3D-grounded accuracy tools enabled image generation with Google and GPT models and video with Seedance. Each asset passed VW review, down to the smallest details.
Car-select shot and in-game race footage from reference photos alone, no CAD. 20+ scenes explored on a lean budget.
Volkswagen's in-house team produced the Tiguan Edition 20 launch stills in days, product-correct down to the wheels.
How Volkswagen shapes the next generation of interactive 3D experiences, with Volkswagen Group Services and Credera.
1Invented badge
2Wrong light bar
3Unreadable plate
4Wrong wheels
General AI image generator
SolidMeta · from the approved CAD
Auto Heal: 0 differences left
Built from your CAD and reference photos, then checked against them frame by frame. What differs is fixed before you approve it.
How every frame gets checked ↓Ask in plain words and attach the photo. The agent picks the skill that fits: Car in new locations.
The plan is a checklist in the chat. The agent stops at Confirm and runs nothing until you click.
Every run lands on the board: the photo wired into a tool node, the result in its own container, one row per location. Open any node and run it again yourself.
Results stay drafts until you keep them. Keep the ones you like, ask for a redo with a note, and the new version lands on the board next to the old one.
From the same chat, the agent prepares a delivery with recipients and their roles. It goes out when you approve it.
Keep the subject exactly as it is and put it somewhere else, described or from a photo.
From one product photo: a clean studio shot, scene variants, every channel's format and a short clip.
Your car photos placed in new surroundings, the car itself untouched, the light matched.
Car photos repaired against a clean photo of the same car, then each checked for what is still wrong, with a PDF report.
The next location and the range come from the product you already approved. Your artist composes the line-up; every car still renders from its own approved CAD. The film uses those same shots.
Campaign stills stay true to the product: body lines, trim, and badge come from your CAD, not a guess.
Your team's Styles: make one from an image, share it, and use it on any generation.
Flux
Gemini
Seedream
GPT
Grok
Qwen
Comment on the picture, see who is on the board, resolve notes in place, and restore any earlier version of the board.
One shot camera, with its move, lens, and framing, drives every video model. Switch between Veo, Seedance, and Wan without describing the shot again.
Fix what's wrong in a clip. Only the bad stretch is re-made and spliced back, and the rest stays. Every character keeps one voice across clips.
Biskuit & Keith — the face, the size, and the voice stay that character’s.
Seedance
Wan
Grok
Flux
LTX
MiniMax
Package the approved stills and motion into a client delivery. Roles decide who can see it, review it, and download it.
CAD (STEP, STP, IGES, CATPart, JT), 3D models (GLB, FBX, OBJ, USD), product photos, or a script.
Image and video models from several providers, among them Flux, Gemini, GPT, Seedance, Veo and Wan. Admins decide which providers and regions each workspace may use.
Every generation lands on a board as nodes you can open and run again. Nothing the agent makes is final until someone keeps it, and Auto Heal reports every fix it made.
Yes. It works with each person's own permissions, asks before it deletes, shares or invites anyone, and checks with you before a large batch. Nothing it makes is final until you keep it. Run it on Claude, Gemini or GPT, or on an open model on your own GPU.
You keep ownership of the content you produce. We don't train AI models on it: not ours, and not anyone else's.
SolidMeta is a product of SolidMeta Inc. in Austin, Texas, and is distributed in Europe by Unevis GmbH. Data of European clients is processed on our servers in the EU. Enterprise plans can run on-prem, air-gapped or in a private cloud.
Yes. Every image and video carries a machine-readable AI marker in its metadata, the one Adobe, Google and Microsoft tools read, and nobody can switch it off. On top of that comes a visible label: the EU AI icons or your workspace's own, and an organisation can make it mandatory for every workspace.
Yes, on Enterprise. A REST API runs generation and Auto Heal, with scoped keys, a cost estimate before each job, and webhooks.
Pierre Grage
CEO · visual effects at Weta Digital, ILM and DNEG
Mirko Hänßgen
CTO · automotive engineering and systems architecture
Alexander Grasse
Product · helped establish VRED in automotive visualisation
Michael Pangrazio
Creative Advisor · ILM, Weta Digital, two Emmys
Danijel Jakovac
Lead Developer · the board and real-time collaboration
Abdallah Huballah
Lead Platform Engineer · native CAD ingestion, from Autodesk and PI‑VR
Studio
Outdoors