While frontier models like Sora excel at short clips, enterprise AI video platforms like Synthesia must build proprietary models. These are essential for creating long-form content and maintaining brand consistency (e.g., logos, backgrounds) across multiple scenes, which consumer-focused models can't yet handle reliably.

Inspired by LLMs, Atlas treats 3D reconstruction as "generation with a really long context." This allows the model to ingest dozens of views to ground its output, bridging the gap between purely imaginative generation and precise, data-driven reconstruction.

While many competitors focus on prompt-based "agentic editing," Tela's founder believes this is a temporary step. The ultimate goal is for AI to analyze a raw recording and automatically produce a high-quality final video without any user prompts or editing commands, leaving only the 'fun part of telling your story'.

The next frontier for visual intelligence is twofold: creating truly multimodal models that retain long-term context of user interactions without re-prompting, and developing real-time generation. Real-time capabilities are crucial for creating duplex interactions and enabling robots to perceive and act instantly.

LLMs excel at 'spatial aesthetics'—arranging elements on a static page. For video, they must learn 'temporal aesthetics,' where information is revealed over time without requiring eye movement. This is a key training challenge for creating compelling AI-generated motion content.

Traditional video models process an entire clip at once, causing delays. Descartes' Mirage model is autoregressive, predicting only the next frame based on the input stream and previously generated frames. This LLM-like approach is what enables its real-time, low-latency performance.

A "world model" transcends simple video generation. It is defined by three key capabilities: real-time responsiveness to user input (e.g., mouse clicks), long-horizon consistency over minutes or hours, and interactivity via multiple modalities like keyboard and voice.

The workflow of generating AI video scene-by-scene and stitching clips together is becoming obsolete. Newer models like Kling 3.0 can interpret multi-scene prompts, creating a single, continuous video with multiple shots. This drastically simplifies production and improves narrative coherence.

The primary challenge in creating stable, real-time autoregressive video is error accumulation. Like early LLMs getting stuck in loops, video models degrade frame-by-frame until the output is useless. Overcoming this compounding error, not just processing speed, is the core research breakthrough required for long-form generation.