elon101-liveportrait.mp4
What’s new in DeepLiveCam 2.8 Ultimate, compared with 2.7.5 Ultimate:
- 400% performance improvement over 2.7.5
- New model — Live portrait has been added for a new experience! Easily create your own VTUBER by just using images! This will go hand in hand with our new
- Optimized HyperSwap processing — works within the face area and reuses memory buffers to reduce unnecessary processing.
- More efficient face enhancement — reduces image-conversion work and data transfers on supported systems.
- Improved frame interpolation — runs on a separate worker and avoids building up delays when processing falls behind.
- Better Windows webcam capture — tries native camera formats to avoid unnecessarily slow capture modes.
- More reliable TensorRT engine loading — checks saved engines against the GPU, runtime and model before using them.
- Updated NVIDIA processing stack — newer CUDA 13.2, PyTorch and ONNX Runtime packages for supported hardware.
- Clearer GPU control — Some UI changes
- Safer video-rendering jobs — validates the settings passed to the rendering process.
Get it exclusively at Deeplivecam.net
Here's a tip! You can use this as a vtuber creator by just generating some images! See demo at https://deeplivecam.net/index.php/news/deeplivecam-2-8-ultimate-early-access