Hacker News new | past | comments | ask | show | jobs | submit login
Doing an HD Remake the Right Way (fortressofdoors.com)
254 points by smacktoward on Sept 29, 2015 | hide | past | favorite | 29 comments



When thinking about what is considered the "original" representation of the sprites, one should be aware of the nature of the original medium (CRT) as well. Most importantly: pixel aspect ratios, color space, overscan ...

Highly recommend reading Kyle Pittman's articles on CRT simulation:

http://www.piratehearts.com/blog/2014/03/28/crt-simulation/ https://news.ycombinator.com/item?id=9364800

http://gamasutra.com/blogs/KylePittman/20150420/241442/CRT_S...


The Rare Replay collection is somewhere else that a lot of work has been put into the CRT emulation for older games.


The art in Skullgirls was done somewhat similar to the way the article describes in "Recoloring". Each frame was composed of three different images and each sprite had a palette along with it. The palette was a list of color gradients. The first image was the black and white line work. The second image was a solid color key layer whose value acted as an index into the palette (i.e. which gradient should be selected). The third image was a monochrome shading layer which acted as the position into that gradient that should be drawn. The gradients were rendered to textures at runtime, and the shader did all the work of actually composing the layers.

It did have some downsides, in particular that it was a bit unnatural for artists to work with, and memory was a constant struggle on console. Skullgirls characters could have well over 1000 frames of animation, and the game supports 3 on 3 fights. Mike did a handful of clever tricks to optimize memory usage, including breaking the images into smaller tiles and throwing out all the empty ones, and compressing the images even in memory. I'm a big fan of the final effect; I think the sprites look great, are high resolution, and allowed artists to create an extremely varied set of palettes for the game.


Hey there! I'm the author of this article. Would love to swap some notes with you sometime about your pipeline! Mind dropping me a line at lars {dot} doucet {at} gmail {dot} com?


Great article.

When an indie developer does an HD remake, it's a labor of love. Each sprite is scaled up meticulously and a lot of thought is put into how to make the game look the best.

When a big company like Squeenix does an HD remake, they toss it to a third party, who will do...well, pretty much what you see here. FFV has never been a big moneymaker for Square (wasn't even originally released in the US, so the nostalgia factor isn't even there for many of us who grew up with their games), so I'm guessing "cheap and good enough" trumped "actually good".

What's funny is that those who are nostalgic for these games (or who just think they're better than most of the stuff on the market today) are willing to work long and hard with no monetary compensation to produce actually good versions of the game in question. I wonder if there's some way for Square to work with these folks to produce a legal, well-done version of the game for Steam. Probably too risky, but it's a shame.


Tracking down rights to rerelease old games is GOG.com's secret sauce. And once they've got the distribution rights, they'll work with people who build emulators (like DOSBox) and wrappers.

See also: http://www.rockpapershotgun.com/2015/09/16/how-gog-com-save-...

Granted, some companies do want to keep those rights to themselves.


Fantastic writeup. Another example I can think of offhand would be the VBA remake of the original Genesis Shining Force.

Some comparisons:

Original: https://www.youtube.com/watch?v=pewTGCkt4qE Remake: https://www.youtube.com/watch?v=3uVuOd4Aiag

Not sure which one is really better - the original has a very different color palette. The sprites are probably nicer on the remake, although I think the character portraits in the original are superior. The remake does have a number of gameplay improvements.


Another interesting example is FF6 Advance. The original GBA had a screen that wasn't backlit, and would be hard to see darker colours in, particularly if played outside or near a window etc. So many games (FF6) included, compensated for this by using a brighter colour palette across the board than NES/SNES/etc versions.

Of course, once you play that version on a emulator on a modern LCD (or even on the backlit GBA SP), the colours end up being too bright, to the point that there is a fanmade patch for FF6 advance to bring back the SNES colour palette:

http://www.fantasyanime.com/finalfantasy/ff6/images/ff6gba_c...


Waifu2X is a powerful upscaling algorithm based on deep convolutional neural networks, and specifically trained on anime.

Huh. I guess it goes to show what good solutions fan groups can come up with. That's fairly bleeding edge, with a well chosen training set and it produces great results. Really fascinating. I wonder how well it works on regular old line-art or if it really is dependent on working with anime-like artwork.


The best part about waifu2x is that it tends to work really well even on images it wasn't intended for, at least in my experience. I use the web version [1] all the time on all sorts of images, usually with excellent results--even for photos.

[1] http://waifu2x.udp.jp/


You could always re-train the network on a different style of images if you have a huge set. The set used with waifu2x was probably easy to get due to pixiv/*boorus.


Reminds me of the HD remake of the Resident Evil remake for the GameCube (the remake of the remake of the PlayStation original, just to be confusing).

Back when it was released on GameCube it looked absolutely amazing. When they released the PC port, more than a decade later, they talked about re-creating some of the iconic scenes in the game engine so that they were more dynamic. In the end only two scenes re-created. The remaining scenes were just upscaled from their GameCube originals. This was unfortunate since not only where they a low resolution (something like 700x300), but some of the scenes were heavily compressed and upscaling them produced a lot of haloing and noticeable compression artefacts.

Why Capcom couldn't just re-render the original scenes is beyond me.


Game companies seem to be particularly bad at retaining original assets. Developers and artists keep a lot of stuff on their locally non backed up machines and regular staff churn, combined with deliberate software obsolescence and bit rot pretty much spreads the ashes of original source files to the cosmos.


I would like to think that is the most likely reason behind why the assets weren't re-rendered. The alternatives are just laziness or budget issues, which I wouldn't put past Capcom either.


The new "Waifu 2X" neural net based up-scaling algorithm is impressive. That should be applied to older anime.

Improving older anime, drawn at low-res and 5 FPS, has real potential. Especially if you can get better source material than old VHS tapes.


The low framerate is a much more obvious, and much harder to solve, problem I think...

You can't easily auto-tween a 3D rotation of a 2D character, for example.

I think the author's approach of "use upsizing algorithms as a basis for humans to remaster" is probably the best of both worlds.


"You can't easily auto-tween a 3D rotation of a 2D character, for example."

Actually, you can. See the previous HN discussion of "Framefree", which decomposes an image into separately moving layers and uses a full mesh morph layer by layer to interpolate between frames. Some ultra-slow motion for sports is done that way.


I assume you mean this? https://news.ycombinator.com/item?id=10052626

That's pretty awesome! I stand by my point that it's not easy, but I'm amazed by how effective that is.


Here is a great interview with the developers of the Gears of War HD remake for Xbox One. They used the original, nine-year-old game engine and scripting to keep the same feel.

http://www.eurogamer.net/articles/digitalfoundry-2015-the-ma...


Right or wrong can only be reckoned with respect to one variable: THE BOTTOM LINE. And Square Enix can make money by turning the crank and shitting out an old Final Fantasy game slapped with a fresh coat of paint, at much lower cost than doing a proper, polished remake and certainly lower than releasing a new game.

I suspect after FF7Remake Square Enix will pull out from console gaming altogether and release more "get Cloud's sword with microtransactions" style mobile farm fests.

Because that's where the money is.


I feel like most (all?) popular art is that way. Games, movies, music.

1. Create something amazing, pour artisan-level craftsmanship into it, gain huge following

2. Exploit your popularity by churning out branded garbage

3. Profit

Even Disney's not immune. Have you seen how many direct-to-DVD/streaming 'movies' there are that cash in on their popular franchises? It sucks.


> Have you seen how many direct-to-DVD/streaming 'movies' there are that cash in on their popular franchises?

I have a strong feeling that a lot of those are made exclusively for toddlers, who tend to frequently get obsessed with consuming anything-and-everything related to a given franchise. I don't feel bad that they're not well-made; nobody could keep pumping good pieces of media out at the rate an obsessive toddler will want to consume them.


My obsessive toddlers would watch the quality stuff over and over again. They didn't need the filler.


Final Fantasy's name is even indicative of that last-ditch artisanship...

> [Sakaguchi's] employer Square refused to give him permission [to make an RPG] as it expected low sales of such a product

> ...the title was later changed to Final Fantasy as the game was thought to be the company's final project under the threat of bankruptcy. Sakaguchi also explained that the new title stemmed from his personal situation: had the game not sold well, he would have quit the games industry and gone back to university.

A bit editorialized on my part, but still.

https://en.wikipedia.org/wiki/Final_Fantasy_%28video_game%29...


Your point is valid, though perhaps misapplied in this case due to nostalgia-tinted glasses. FFV is a tremendously fun game, but I don't think it's accurate to say it received artisan-level craftsmanship. I don't think it was a transparent money grab the way the HD remake is, but compared to the ones that came before and after, it's...kind of the Godfather 3 of the SNES era. Plenty good on its own, but heavily overshadowed by the others bearing its name.


Maybe, but anecdotally, I saw the FFV remake on steam, and I'm the target audience and have bought a number of versions of various Final Fantasy games and remakes. However, I didn't buy the FFV remake because it looked horrible versus the standard def version.


Didn't FFV already get an amazingly well done remake on the DS? Or was it FFIV?


FFIV had a remake with poor 3D graphics on DS. It later had a remake with excellent 2D graphics on PSP.


shame on you! the 3d remake is amazing - pixels are a good thing when you do it right.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: