I spent all weekend trying to make backgrounds look like something other than a mess

Watching the pixels jitter while trying to work

I sat down on Saturday morning with a cup of lukewarm coffee, determined to finally finish the background for this webtoon chapter. I had been using these stock vector images I downloaded from a site that claimed to be royalty-free, but honestly, the more I looked at them, the more they felt like they were fighting my character art. Every time I tried to trace over them or even just use them as a loose guide, something felt off. The perspective was always just a degree or two away from matching my own. I kept adjusting the lines, overthinking the vanishing point, and by noon, I had barely managed to sketch a single alleyway. I kept thinking about how much easier it would be if I could just let the software handle the heavy lifting of lighting and texture without it looking like a weirdly polished filter.

The endless promise of new tech

I saw some news earlier about Nvidia pushing this new DLSS 5 stuff, talking about real-time neural rendering and how it handles motion vectors to make light bounce off fabrics or skin realistically. It sounds impressive on paper. When I read about it, I half-wished my outdated PC could actually render something that looked that clean. Instead, I’m stuck here manually moving vector nodes, trying to make a brick wall look like it isn’t just a flat, lifeless object. I found myself looking at the price of upgrading my graphics card, but the thought of dropping 800 dollars or more just to see some better reflections in my viewport felt ridiculous. It’s not like my readers are going to notice if the texture of the pavement has real-time light scattering. They just want to see the characters talk.

Why manual tracing still feels safer

I ended up going back to that site—I think it was one of those resources for webtoon creators where you can buy access for around 15 or 20 dollars a month—and looked for something simpler. There’s a specific kind of frustration that comes with using automated tools that try to guess what you want. If I use an AI-assisted tool to generate the base, it always feels like the character doesn’t quite belong in the room. I keep comparing it to just drawing it by hand, which takes three times as long but at least doesn’t give me that uncanny valley feeling. I remember a friend told me they just use photos of real locations and trace them. It sounds practical, but even then, you’re stuck spending hours deciding which details to keep and which to erase so it doesn’t look like you just pasted a photo onto a canvas.

Getting stuck in the weeds

Around 4 PM, I realized I had spent two hours just trying to get the shading on a tree right. The motion data and light paths that these new engines talk about seem like magic, but in practice, I’m just hitting ‘undo’ every few seconds. I tried to see if there were any ‘no-copyright’ background sites that had cleaner lines, but most of them either looked too cartoony or too complex. It’s an weird irony that the more advanced these rendering technologies get, the more I feel like I’m just struggling with the basic task of making a drawing look finished. I could have probably just drawn a few simple shapes and called it a day, but I’m too deep into this now.

Why I keep doing this to myself

I still haven’t settled on the right workflow. Every time a new update comes out for a drawing app or a new plugin promises to handle perspective grids or vector tracing automatically, I get excited for a second, then realize it’s just another thing to learn. It feels like I’m constantly chasing a version of my art that looks professional without actually having to put in the decades of training. I saved my progress, closed the laptop, and went to the kitchen to find something to eat. I’ll probably wake up tomorrow and change the entire background anyway, because looking at it again with fresh eyes usually reveals that I did something completely wrong.