Category assignments and random selection counts for watermarks and scrolling text, multi-GPU transcoding, and improved GPU utilization
This update revisits the GPU transcoding logic, which had not received much attention for a while. It now supports multiple graphics cards: however many your server has, it uses them all. Each new job is assigned to the least-utilized card. New jobs stop starting when the memory usage of every card is nearly 100%, effectively maximizing utilization. We also made adjustments to scrolling text and watermarks.
Changelog
- Added resolution, bitrate and frame-rate settings for segment ads, so their parameters can be adjusted.
GPUtranscoding now supports animated-image watermarks and simultaneous processing across multipleGPUcards, greatly improving utilization. Each new job automatically uses the least-utilized card. New jobs pause when every card's memory usage approaches100%, then resume when memory becomes available.- Added category assignments for watermarks and scrolling text. When multiple watermarks or scrolling-text items are configured and assigned to categories, they apply only to the corresponding category.
- Added random selection of watermarks and scrolling text. Set the desired number separately for each type, and every transcoding job randomly uses 1 or more image watermarks or scrolling-text items.
- Added horizontal mirroring. When enabled, the transcoded picture is mirrored horizontally, which can correct certain short-video camera issues or make random watermarks and text in a video less recognizable.
- Fast slicing can generate directly downloadable
mp4URLs, processing the source video by default. Normal transcoding continues to use the transcodedMP4for this purpose. - Fixed image encryption also affecting images that disguise
TSsegments, causing playback failures. - Improved image-to-
txtconversion. When enabled, the original image displays no content; changing the URL extension fromjpgtotxtconverts it automatically totxt. - Added
Azure Blob Storageto cloud-storage synchronization.
Resource-usage screenshot of a server with two graphics cards:

The memory usage of both 2 graphics cards is nearly at its limit simultaneously.


