Faceswap 1.2.0 — Ai Video

Unlike many web-based SaaS tools, version 1.2.0 keeps everything on your local SSD. No uploads to a central server, no hidden watermarks, and no monthly subscription.

For editors working on music videos or dialogue scenes with multiple characters, this is a game-changer. The new batch processing pipeline allows users to define a swap map (e.g., "Actor A becomes Face 1, Actor B becomes Face 2") for an entire video sequence. The AI analyzes facial landmarks across all subjects simultaneously, reducing multi-face swap render times by nearly 70% compared to sequential processing.

Previous versions struggled with real-time applications unless running on enterprise-grade A100 GPUs. AI Video FaceSwap 1.2.0 introduces a pruned MobileNetV4 backbone fused with a StyleGAN3 decoder. The result? A consistent sub-100 millisecond inference time on consumer hardware (RTX 3060 and above). Live streaming at 60fps is no longer a pipe dream; it is the default baseline.