As part of the ongoing Wing Commander IV Resmastered project (https://wcrespace.com), I've upscaled and remastered all the game's rooms, environments and sprite animations using ESRgan AI/Deep Learning. These are a few samples, along with comparisons to the original 640x480 room visuals.
You might think that remastering the rooms and sprite animations for Wing Commander IV would be a similar task to remastering the videos – I certainly did before I started – but it’s actually a very different beast.As previously mentioned, the team is focusing on WC4, but occasionally they gain new insights with experimenting in WC3. Here's an upgraded gameflow demo Pedro recorded as well:
For a start, all the “shapes” (as the game files call them) are static images. I was expecting that for the room backgrounds, but not for the sprite animations (Eisen and Maniac on the bridge, Vagabond in the bar etc.). Yet when Pedro extracted the files from the original game and sent them my way, there they were – eight thousand, five hundred and thirteen TGA files, comprising the rooms, hangars, computer consoles, buttons, toggles and every individual frame for every character animation.
Upscaling static images using AI is quite different to upscaling video footage. When you have an image sequence in a video file, the AI can analyse groups of frames, identify similarities and changes between them, gather and analyse depth information from parallax motion and interpolate a lot of data to add to the image. With a static frame, you don’t get that.