A San Francisco startup thinks the solution to generative AI video's biggest limitation isn't better prompts—it's spatial intelligence that lets creators control scenes like a 3D film set.

Intangible's web-based platform combines traditional 3D scene building with generative AI rendering, allowing users to position objects, set cameras, and direct scenes before AI brings them to life. The approach addresses a fundamental challenge: current text-to-video tools rely on language interpretation that often produces unpredictable results, making them unsuitable for professional workflows. By grounding AI models in structured 3D environments, Intangible aims to deliver the precision and control that filmmakers and content creators actually need.

Founded by Charles Migos and Bharat Vasan with $4 million in funding from a16z Speedrun and other investors, the company represents a fresh approach to an industry problem that has frustrated creators since AI video tools first emerged.

Scene Control: Moving Beyond Text Prompts

Current AI video generation feels like rolling dice. You describe what you want, cross your fingers, and hope the AI interprets your vision correctly. This fundamental limitation has kept these tools relegated to memes and experimental content rather than professional workflows.

Intangible's solution centers on what computer vision expert Fei-Fei Li calls "spatial intelligence"—the ability to perceive, interpret, and interact with three-dimensional space. Instead of describing a scene in words, users build it directly using a drag-and-drop interface with over 5,000 preset assets.

"Current AI models are reliant on extensive prompting, and language alone isn't enough to convey creative intent," explains Intangible chief product officer Charles Migos. "By providing generative AI models with spatial intelligence, Intangible allows creatives to get closer to professional-grade results with less prompting, more feel, and more control."

The platform's interface resembles familiar 3D software but simplified for non-technical users. You start with a preset scene or blank environment, then add buildings, characters, and objects from the asset library. Standard 3D controls let you move, scale, and rotate elements precisely where you want them.

Camera Direction: Bringing Film Set Logic to AI

The platform's strength lies in its camera system. Users can position multiple cameras throughout their 3D scene, set up movement paths using waypoints, and control exactly how each shot unfolds. This mimics traditional filmmaking workflow where directors plan every angle before rolling camera.

Once your scene is staged, Intangible's AI rendering engine interprets the 3D layout and generates photorealistic video. The system currently uses Kling, a Chinese AI video generator, to transform the composed scenes into final footage. This approach provides significantly more predictable results than pure text prompting.

"By building in interactive 3D from the outset, Intangible's world model gives generative AI image and video generation models the ability to be more precise, without extensive prompting," says CEO Bharat Vasan.

The workflow addresses what Migos calls the "interpretable nature of language" problem. When you describe a scene in words, every person—and every AI—visualizes it differently. That ambiguity works fine for novels but creates chaos in video production where precision matters.

Professional Pipeline: Targeting Real Production Needs

Intangible positions itself specifically for professional creators rather than casual users. The founding team includes veterans from Unity, ILM, Electronic Arts, and other major studios who understand production workflows intimately.

Their target applications span multiple industries:

  • Film previsualization - Directors can sketch entire sequences in days rather than weeks

  • Advertising pitches - Agencies can create interactive concept presentations for clients

  • Marketing activations - Brands can prototype virtual events and experiences

  • Game development - Studios can rapidly prototype environments and scenarios

The platform's web-based nature removes technical barriers that typically limit 3D tools to specialists. No downloads, installations, or high-end hardware required—just a browser and creative vision.

However, early testing reveals the system's current limitations. The Fast Company review found rendering inconsistencies and historical accuracy issues when testing with Roman-themed scenes. The interface also experiences performance sluggishness, particularly during initial sessions.

Market Timing: Filling the Professional Gap

The timing reflects broader industry frustration with current AI video tools. While platforms like Runway and Luma have advanced significantly in visual quality, they remain fundamentally unpredictable for professional use. The disconnect between creative vision and AI output has kept these tools on the sidelines of serious production.

Intangible's approach directly addresses this gap by putting spatial control first. Rather than trying to improve text interpretation, they're changing the input method entirely. This strategy could prove more sustainable than competing on prompt engineering alone.

The company's $4 million funding round signals investor confidence in this approach. Backers include experienced technology investors who understand both the creative industries and AI development challenges.

Technical Foundation: Building on Established Principles

The platform's architecture combines proven 3D rendering concepts with modern AI capabilities. Users work within familiar paradigms—cameras, lights, objects, and scenes—while AI handles the complex task of photorealistic rendering.

This hybrid approach offers several advantages:

  • Predictable results - 3D positioning removes spatial ambiguity from AI interpretation

  • Iterative refinement - Users can adjust scenes and re-render until satisfied

  • Professional compatibility - Output integrates with existing production pipelines

  • Collaborative workflows - Teams can share and modify 3D scenes directly

The system's reliance on external AI engines like Kling represents both strength and limitation. While it allows Intangible to focus on interface innovation rather than AI model development, it also means they're dependent on third-party rendering quality.

Development Roadmap: Expanding Creative Control

Intangible has outlined ambitious expansion plans extending beyond their current capabilities. Planned enhancements include more granular editing tools, enhanced asset compatibility, and collaborative features for team-based projects.

The company also expects to expand into adjacent markets like architecture, education, and live event production. These sectors share similar needs for spatial visualization and could benefit from accessible 3D creation tools.

"In the next three years, we expect tools like Intangible will be able to cover all aspects of preproduction and digital production for existing forms of media," Migos and Vasan explain. They envision their platform enabling new creative categories that blend linear, interactive, and immersive media.

Final Cut: Redefining Creative Control in AI Video

Intangible's approach represents more than technical innovation—it's a fundamental rethinking of how creators should interact with AI systems. By prioritizing spatial intelligence over language interpretation, they're addressing the core limitation that has prevented AI video tools from entering professional workflows.

The platform's current limitations don't diminish its conceptual strength. Performance issues and rendering inconsistencies are solvable problems, while the underlying interface philosophy offers a sustainable path forward for AI video creation.

As the technology matures, Intangible could establish the standard for how creative professionals control AI-generated content. Their emphasis on human-centered design and professional workflow integration positions them well for an industry ready to move beyond experimental applications toward practical production tools.

Reply

or to participate

Keep Reading

No posts found