Inside Our Latest AI Model Release
This release focused on three things creators told us mattered most: speed, texture fidelity, and lighting that feels human.
Faster Rendering
We rebuilt our inference pipeline to cut average render time significantly, so you can iterate on ideas without losing momentum. Faster feedback loops mean more experimentation, and more experimentation means better final results.
Richer Textures
Fabric, skin, foliage, and metal now hold up to close inspection. The model was trained on a broader and more carefully curated dataset of high-resolution reference imagery, which shows up most clearly in macro details — the kind that used to give AI images away at a glance.
More Human-Like Lighting
Lighting is one of the hardest things to get right because it has to be physically consistent across an entire scene. This update improves how the model reasons about light sources, shadows, and reflections, producing images with a more natural sense of depth.
What's Next
This is the first of several planned updates this year. We're already working on improvements to consistency across multi-image sequences and finer control over composition. Stay tuned for more.