Since about a year ago, FFmpeg has supported the new libvmaf API and features, the most significant of which is that the VMAF models are now included in the library itself. So there's no need to point to model files anymore, unless you're using a self-trained/third-party one. So for example, specifying the 4K model is simply done with libvmaf=model=version=vmaf_4k_v0.6.1. Currently to use anything other than the default model with Av1an, you need to download the model in question and use --vmaf-path to specify the model file.
See also #659 and #488.
Also, it would be useful if Av1an used the 4K VMAF model by default for output files significantly larger than 1920x1080. The default model is trained to predict quality on a 1080p HDTV, and for anything larger than that the 4K model should probably be used. av-av1 does this by default (for resolutions larger than 1440p I believe).
Since about a year ago, FFmpeg has supported the new libvmaf API and features, the most significant of which is that the VMAF models are now included in the library itself. So there's no need to point to model files anymore, unless you're using a self-trained/third-party one. So for example, specifying the 4K model is simply done with
libvmaf=model=version=vmaf_4k_v0.6.1. Currently to use anything other than the default model with Av1an, you need to download the model in question and use --vmaf-path to specify the model file.See also #659 and #488.
Also, it would be useful if Av1an used the 4K VMAF model by default for output files significantly larger than 1920x1080. The default model is trained to predict quality on a 1080p HDTV, and for anything larger than that the 4K model should probably be used. av-av1 does this by default (for resolutions larger than 1440p I believe).