I’m utterly terrified of them pulling an Andromeda/2077 and getting stuck in dev hell trying to debug the new engine bits instead of actually building the game. This is the advantage of prebuilt engines: someone else has already one all the hardware support/optimization and contemporary architecture stuff for you.
I’m less afraid of them pulling a Starfield, I suppose. The “divinity engine” in BG3 already runs okay. It’s not sleek like CryEngine KCD2, but it doesn’t feel janky or dated either, and even the mildest refresh over BG3 would be fine.
Much less is determined by engine than the average person thinks. Andromeda wasn’t a new engine; it was an engine that was made to make Battlefield games that then had to be used to make action RPGs and racing games after the fact. Capcom made an engine for the games they had in mind 10 years ago, and it’s fantastic at Resident Evil, Devil May Cry, and even serving as an emulation wrapper, but it’s showing cracks under the support for open world games that they added more recently. Larian’s engine is made to support the systems driven RPGs they conceptualized in the early 2010s, and there’s little chance some other engine will do it just as well or better without plenty of custom code anyway. Ask Digital Foundry about all of the “optimization” Unreal 5 has done for developers already.
If they can shoehorn in something akin to KCD2’s or Satisfactory’s Global Illumination, but keep their dev workflows and existing systems in place, that’d be perfect.
From the other Larian article in this community, it seems their engine improvements are largely things that they claim will allow them to iterate on ideas faster, like going right from mocap to a usable animation more quickly.
I truly love that Larian leadership frames everything they talk about around devs and their needs/wants. Another D&D game? “Oh, that’s great and all, but our devs hearts weren’t in it so we dropped it like a rock.” New engine? They ramble about improvements to dev workflows. It is so obviously a top priority.
Keep in mind that also comes with Vincke championing AI, and though he says no genAI assets will make it to the final product, there’s still some dissent. Here’s hoping though.
Well that can be reasonable. Obviously don’t vibe code an engine, but LLMs are great for basic code autocomplete, or quick utility scripts, things like that.
Really specialized AI (not LLMs/GenAI) can be great at, say, turning raw mocap into character animations. Or turning artist sketches into 3D models. Cogs in their pipeline, so to speak, which has nothing to do with GenAI slop making it into a final product.
The line is very fine though, and most in the business world skew to the side of pushing slop.
I think very little about AI compared to most people, for or against, but it largely seems to me like a solution in search of a problem, and it’s very cult-like how many CEOs get on board with it so quickly despite its very public lack of actually good results. On paper, the way Vincke describes their use of it sounds fine to me, but hopefully he’s not doing something so idiotic as to mandate its usage, as is happening at workplaces for friends of mine right now.
“Mandate its usage” could mean the motion cap/animation people have to learn some kind of automation tool, that’s now part of the engine.
That’s fine.
And that’s very different from the “you MUST make X hits to Microsoft Copilot” type garbage that’s so common now.
I’m harping on this because I’m afraid Larian will try something reasonable, yet get immense, unwarranted backlash (both internally and publically) because of other workers’ experiences with enshittified ML. And how politicized “AI” is becoming.
Machine learning is not bad. Tech Bro evangelism and the virus they spread among executives, is. And I don’t want the garbage they sell to poison tools studios like Larian could use to get ahead of AAA publishers.
Depends how much they “redo”.
I’m utterly terrified of them pulling an Andromeda/2077 and getting stuck in dev hell trying to debug the new engine bits instead of actually building the game. This is the advantage of prebuilt engines: someone else has already one all the hardware support/optimization and contemporary architecture stuff for you.
I’m less afraid of them pulling a Starfield, I suppose. The “divinity engine” in BG3 already runs okay. It’s not sleek like CryEngine KCD2, but it doesn’t feel janky or dated either, and even the mildest refresh over BG3 would be fine.
Much less is determined by engine than the average person thinks. Andromeda wasn’t a new engine; it was an engine that was made to make Battlefield games that then had to be used to make action RPGs and racing games after the fact. Capcom made an engine for the games they had in mind 10 years ago, and it’s fantastic at Resident Evil, Devil May Cry, and even serving as an emulation wrapper, but it’s showing cracks under the support for open world games that they added more recently. Larian’s engine is made to support the systems driven RPGs they conceptualized in the early 2010s, and there’s little chance some other engine will do it just as well or better without plenty of custom code anyway. Ask Digital Foundry about all of the “optimization” Unreal 5 has done for developers already.
This is a fair point. When I made the original comment, I didn’t realize their in house engine went so far back:
https://en.wikipedia.org/wiki/Divinity_Engine
If they can shoehorn in something akin to KCD2’s or Satisfactory’s Global Illumination, but keep their dev workflows and existing systems in place, that’d be perfect.
From the other Larian article in this community, it seems their engine improvements are largely things that they claim will allow them to iterate on ideas faster, like going right from mocap to a usable animation more quickly.
That sounds excellent.
I truly love that Larian leadership frames everything they talk about around devs and their needs/wants. Another D&D game? “Oh, that’s great and all, but our devs hearts weren’t in it so we dropped it like a rock.” New engine? They ramble about improvements to dev workflows. It is so obviously a top priority.
Keep in mind that also comes with Vincke championing AI, and though he says no genAI assets will make it to the final product, there’s still some dissent. Here’s hoping though.
Well that can be reasonable. Obviously don’t vibe code an engine, but LLMs are great for basic code autocomplete, or quick utility scripts, things like that.
Really specialized AI (not LLMs/GenAI) can be great at, say, turning raw mocap into character animations. Or turning artist sketches into 3D models. Cogs in their pipeline, so to speak, which has nothing to do with GenAI slop making it into a final product.
The line is very fine though, and most in the business world skew to the side of pushing slop.
I think very little about AI compared to most people, for or against, but it largely seems to me like a solution in search of a problem, and it’s very cult-like how many CEOs get on board with it so quickly despite its very public lack of actually good results. On paper, the way Vincke describes their use of it sounds fine to me, but hopefully he’s not doing something so idiotic as to mandate its usage, as is happening at workplaces for friends of mine right now.
Well again, it depends.
“Mandate its usage” could mean the motion cap/animation people have to learn some kind of automation tool, that’s now part of the engine.
That’s fine.
And that’s very different from the “you MUST make X hits to Microsoft Copilot” type garbage that’s so common now.
I’m harping on this because I’m afraid Larian will try something reasonable, yet get immense, unwarranted backlash (both internally and publically) because of other workers’ experiences with enshittified ML. And how politicized “AI” is becoming.
Machine learning is not bad. Tech Bro evangelism and the virus they spread among executives, is. And I don’t want the garbage they sell to poison tools studios like Larian could use to get ahead of AAA publishers.