Meet DAWN, a new AI VFX foundation model from British startup PXLD that composites real actors into entirely different environments while preserving their original performances.
No AI-generated faces. No rewritten expressions. No replacing the actor with a synthetic version.
DAWN introduces a different approach to AI compositing: world harmonisation.
Instead of regenerating an entire video, the model is designed to preserve the original footage while adapting lighting, shadows, reflections, and the surrounding environment.
Key capabilities demonstrated and announced by PXLD:
• Hair-level detail: preserving individual hair strands, fine edges, and subtle facial expressions.
• Performance preservation: keeping the actor's original identity, movement, and facial performance intact.
• Dynamic AI relighting: adapting illumination, shadows, and reflections to match a new environment.
• Flexible environments: placing real footage into photographs, videos, 360° panoramas, and 3D scenes.
• Spatial lighting control: positioning light sources around the scene without relying on text prompts.
• Film-scale ambitions: PXLD claims support for unlimited clip lengths and workflows ranging from smartphone footage to high-end cinema production.
For compositors working in Nuke, After Effects, or professional VFX pipelines, the implications could be significant.
Less manual masking. More control over lighting. And most importantly, the original performance stays the original performance.
DAWN is currently being introduced through early access. Its production capabilities and limitations still need independent validation.
Explore DAWN and join the waitlist:
https://pxld.ai
Could this change the future of VFX compositing?
#DAWNai #VFX #AICompositing
YouTube / ABVX
