Guiding AI with 3D

AI images and videos look great, until you try to make more than one. Then characters change, lighting shifts, and few things stays consistent.

3D can make a big difference here. It locks in camera angles, composition, lighting, and character details, ensuring the AI generates consistent images and videos.

It also works for multiple subjects and longer videos. I created a 30-second story-driven game trailer. No post-production, just 3D keeping the AI in check.

‘Saboteurs’ - a made up story teaser for a deceptive game. You’re right, she is the culprit.

Below I’ll show you how I used 3D to make AI generations more predictable, and how 3D helps you get creative.

3D gives you creative advantages

Test, tweak and explore different art styles

Switching art styles used to mean rebuilding everything from scratch. With 3D you can turn a scene from anime to a paper cutout to anything you want, without losing your design.

Customise characters by attaching anything

In 3D every accessory and every tweak, is fully under your control. Small details can make a big difference. I simply added a baseball cap and a cigarette model for a slightly melancholic mood.

Mix styles in the same scene

3D makes it possible to mix styles in the same scene with control. Like an anime character in a photorealistic world.

Consistency across multiple scenes

Training AI models for consistent subjects

In 3D, any character or asset can be separated and trained as a subject. This is done by capturing multiple angles, expressions, and details. Once trained, it carries over into any scene without losing proportions, details, or style. You can also do this for backgrounds and environments.

Get proportions right, and no more floating objects

In a 3D scene, depth is precise, not a rough guess. A detailed depth map keep proportions, placement, and scale consistent across shots. This makes it less likely for AI to distort objects, misplace elements, or create floating artefacts.

Experiment: a way to get perfect animation consistency

This frame-by-frame generation workflow could enable a new way of working with AI where you have complete control over animations. And you could re-generate any broken frame to make a completely fault free video. Work in progress.

Wrapping up

Make a 3D scene in minutes without any experience

You don’t need to model anything from scratch. Drag and drop an image and a 3D model is created. Or type a prompt. Its fast and easy.

When you’re done just hit render, and everything else happens automatically. The tool refines and relights your image. With these you then simply generate videos.

Is 3D guided AI right for you?

There’s other ways than 3D to get control that are best at some use cases, but if you need full creative control and consistency across a several shots a 3D-first workflow might be the best solution. Essentially for creators who want AI to work with their process, not to take it over.

So, if you are someone who want to promote your game, make animated movies, or create stylised art, then you’d most definitely benefit from 3D guided AI.

Want a chance to test the free beta?

If you want to finally try to make that image or video you simply can’t do today, then signup to the private beta wishlist below.

About us

We’re two founders with former startup experience from building an AI game maker, a multiplayer web based game engine, and more. When trying to make these products we firsthand experienced the issues with AI for code, image, and 3D generation. But we also realised that there’s big benefits with letting 3D guide AI, and we’re now exploring that space.

If you’re curious about learning more about the company we’re building then reach out at tobias@backdrop.tech or join our Discord.