I prefer “list” view over grid view. Switching to grid view shows six videos before the break, but significantly less information about the video.
I prefer “list” view over grid view. Switching to grid view shows six videos before the break, but significantly less information about the video.
What if we host the videos at 144p then have the client use ai upscaling to 4k?
No matter how good an AI is, it cannot restore details that were lost. It can approximate them, but if you have a 4k photo of a piece of paper with a small stick figure drawn on it and compress it to 144p, you will have a gray blob in its place at best or just nothing.
The most advanced AI from 50 years into the future would still not be able to restore the original stick figure.
I think you could do this if the compression algo is made with ai upscaling in mind. You’d just neet to store the word “stick figure” and its position, and bam your ai can draw a stick figure on the sheet.
You store what objects are seen in text or tokens and the pixels serve as more of a coordinate system to mark size/shape etc
I feel like upscaling from 144p would be terrible. 360p might be more doable i think
hmm you’re right. What if we ditch the video entirely and they just enter a prompt of what the video should be and we generate it. With AI?