While Video Enhance AI has worked on Apple’s new integrated M1 System on a Chip architecture using Rosetta emulation, we’re now excited to offer native support for the latest generation of Apple M1, M1 Pro, and M1 Max computers. Apple M1 users can expect a 20-30% performance boost across the board when compared to using Rosetta emulation, as well as significant gains when compared to Intel-equivalent machines. Here is a chart illustrating the performance gains across several Apple computer configurations.
We introduced the Chronos model to provide exceptional quality when slowing footage down (up to 2000%), as well as converting the frame rate of your video clips. Both of these frame interpolation processes can be very processor-intensive and could take a long time to complete. The new Chronos Fast model addresses this speed issue by analyzing larger blocks in each frame, which is especially helpful for high-resolution videos beyond HD (think 4K and 8K clips). As a result, the new Chronos Fast model provides an impressive 2-3x faster processing. Here are two charts illustrating the processing speed increase of Chronos Fast on both Mac and Windows machines, as well as a clip slowed down by 400% while also increasing the frame rate to 60 FPS.
One of our chief goals when training updates to our AI models is to provide faster processing times and improve fine-tuned details for better frame quality. This updated training helps to reduce artifacts and improve details when compared to previous versions. That is why we have optimized new Proteus, Artemis, and Dione models, all of which now deliver 20-30% faster performance for Windows-based GPU-enabled machines. We are also working hard to bring similar performance improvements to our Mac users soon.
We’ve tightened up a lot of usability issues to make Video Enhance AI more stable and reliable. You will now be provided with a much more accurate ETA estimate when performing video processes in the timeline. We’ve also improved the consistency of maintaining model parameters when using batch processing, along with a number of other functional improvements.