Six mistakes I made in my dioramas-and-felt Steam game and one I didn't | Hacker News Reader

Six mistakes I made in my dioramas-and-felt Steam game and one I didn't

novalis.org

81 pointsmjd41 comments

The Dream Machine from Cockroach and of course The Neverhood from Doug TenNapel are some fantastic prior art here. TenNapel even reused his skills to make Armikrog two decades later.

I wonder if the creators of these physically rendered games ever get together to talk shop, or even host a conference, on the topic of non-linear animation and storytelling. It’s the sort of thing that I, as someone who merely consumes their content and is absolutely nowhere near being able to replicate their skill, would love to take part in.

Comic-Con but for games done the hard way. Thimbleweed Park, another remake in the animated adventure game genre, actually has a scene in it that’s based on a conference for real-life point-and-click games developers.

I'm so glad somebody else has heard of The Dream Machine, it's great!

Harold Halibut is also worth a look, though it's not fully stop-motion I believe

Harold Halibut isn't stop motion at all; they made physical puppets and then scanned them into Unity and tried to replicate the look.

It's a gorgeous game, but I was really disappointed in the gameplay. I wouldn't even call it a game; it's basically a movie where you sometimes interact with things, but everything is scripted.

Ahhh, sad to hear that, I only played the demo

Hylics by Mason Lindroth is the only other thing I'd ever seen do the "physical art" thing in a videogame. I love it a lot, so now curious about the other games you've mentioned! Thanks for the recs.

With respect to Doug TenNapel’s contribution to the field, Armikrog is only of interest I think to die-hard nostalgia fans. Maybe one day it will be (or already has been?) remastered into an even better game, but when I played it on release it was more like a mise en place of a great meal than a great meal itself.

The Dream Machine is all out body horror excellence from an indie studio in the truest sense. The fact that they saw the project through to the end, in real time, over the course of several years and while communicating the whole time with their Kickstarter backers is just as impressive as the game itself. Pure brilliance.

There's only a demo out so far, but Scarlet Deer Inn uses embroidery for its character images.

https://store.steampowered.com/app/1553260/Scarlet_Deer_Inn/

Is it multi-threaded?

Two disjoint comments:

1. Film sets contain wild walls that can be easily removed to accommodate light and camera placement outside the space looking inwards. Wild walls in the dioramas could have prevented the necessity of cutting holes.

2. Git is not the right SCM for the job. My first choice would have been Plastic a few years ago, but this got rolled into Unity DevOps after an acquisition. But it is meant to accommodate truly vast repositories (53GB isn't actually that big). The second best choice is SVN - yes, really.

I find it funny how often people look down on SVN outside of game development. It really is a great tool for that job.

The trunk our of studio's SVN is hundreds of GB and the full repo is many TB. And SVN handles this no problem.

What is it that SVN does particularly well here? Handle binaries?

Yes, SVN handles large amounts of big binary rather efficiently. It employs some binary diffing internally to keep storage usage low.

The centralized model also allows optional file locking, which reduces the risk of accidental merges destroying a file. This is a major concern for files that aren't source code when no dedicated merge tool exists.

Also, the centralized storage of the repo history becomes a bit of a bonus at this scale as it doesn't need to be replicated to each client, saving a little bit on disk storage and data transfers.

The one unfortunate downside of SVN is the implementation of merging between branches. That relies on per-file merge histories that are partially stored in per-file properties. This is opaque, internal information, yet it is tracked the same way as file changes. This leads to situations where it's all too easy to discard some of the merge history with harmless looking operations like reverting a file between a local merge and its merge commit to get rid of an unwanted change. Once something like this happens, future merges end up doing random looking things like re-applying changes that have already been merged. This is a major part of what gives SVN its bad reputation. You really have to tread on eggshells when merging. If you can avoid the pitfalls, merging works well. But it requires rather deep knowledge of SVN and discipline.

With LFS handling assets with Git shouldn't be a problem. From the sounds of it, the biggest problem they had here was not in the technology but GitLabs pricing which doesn't separate LFS storage. For example GitHub and BitBucket have separate pricing for LFS, both at more reasonable $.1/GB, i.e. five times cheaper than GitLab. If I'm reading it right Azure even offers LFS for free(!).

Technically I believe you can even have LFS hosted in different place than the repo itself, although I don't know if these commercial providers handle such case. But in theory you could host your git repo at gitlab.com and selfhost lfs if you want to penny pinch.

Git-annex is another good option: https://git-annex.branchable.com/

I've been using it for many years now for data repos like photos. IIUC one advantage it has over LFS is its model makes it easier to store data in many different places (though I don't use LFS myself).

The one standout quality that git LFS has in my experience is the rate at which working copies with LFS enabled shred themselves and become unrecoverable. I've seen this on multiple projects and different users now and it doesn't give me the confidence I'd need to rely on it for a big project.

I'm sure that there will be people replying that this is anecdotal etc. But this is my personal experience and what shapes my recommendations.

Yeah, this plus the apparent lack-of-planning regarding lenses & field of view make me wonder if OP had any of the background they should have had in stop-motion animation?

Are there any prior examples for AI interpolation in stop-motion animation?

This is the sort of project that, no matter how much you know, or think you know, and no matter how much experience you had, it'll still bring you right back down to Earth and humble you somehow. There is no curriculum for preparing for this. There is only "jump in and figure it out as you go".

So let's not criticize anyone for jumping in and figuring it out as they go, as it is the only way to produce a project like this.

Walnut burl veneer is inevitably both warped and also brittle.

If you're going to use it, it's worth making up or buying veneer flattening solution, wetting the veneer with it, and clamping it between two pieces of melamine or plywood with some packing paper to get it flat without destroying it. Change the paper a couple times as the veneer dries.

Once it's flattened, you can glue it down using a caul (one of the pieces of melamine you used to flatten it will work) to hold the veneer flat to your hopefully flat substrate while the glue dries, or you can hammer veneer it with hide glue.

Hammer veneering doesn't risk gluing your caul to your veneer, but it does require a warmer shop, and I wouldn't want to do e.g. a dining table that way. Melamine cauls usually pop free pretty readily if they do get glued to the veneer. If you don't have or want to buy melamine, a packing paper layer between the veneer and the caul will help, as will covering the caul with packing tape.

Ideally, you don't end up with a lot of glue bleeding through the veneer, but burl veneer tends to wick it through.

> But it would have been a mistake to build High Mountain Abbey in Rust.

I didn’t expect avoiding the custom engine side quest in a game with a very custom art style; kudos!

I think there may be an even bigger mistake avoided here: many indie devs build amazing projects and don’t talk about them enough. If you ever created something you enjoyed, wanted to share it and somehow didn’t, watch and learn; I know I should!

> So what I should have done is: first, create the Steam page for the game, then post the first blog post. Well, better late than never: please wishlist the game.

In case the author is reading: anywhere that's not Steam to wishlist or buy the game? This would be a definite purchase from me on GOG - well, if it reviews reasonably well. I love quirky and unusual graphical styles.

But I have never, and probably will never, buy game from anywhere that requires me to install a launcher and look at advertising before I get to play.

>But I have never, and probably will never, buy game from anywhere that requires me to install a launcher and look at advertising before I get to play.

You'll be happy to hear that Steam does not require you to do any of that. If the game does not heavily rely on Steam services you don't have to have Steam started at all. I just tested some games and i could just start them from the executable without Steam running. I also disabled the news window you can get on startup and the Steam client starts on the library, not the store page.

So have fun.

Some games on Steam are truly "DRM-free", but the default way is to use the Steamworks DLL which tries to open the game through Steam if the exe was clicked directly. This "DRM" is trivial to remove with the right tools (if not accompanied by a real DRM), but still you'd be basically cracking the games you bought.

GOG (and also Itch and sometimes Humble) simply gives you an installer with no DRM at all, meaning that the games will continue to work even if GOG blows up.

So yeah, that's why some people prefer buying on GOG over Steam. That, and European patriotism, sometimes.

Creators are sometimes afraid of this model, sadly, due to the perceived increased potential of piracy. Even though CD PROJEKT RED have repeatedly demonstrated that they can sell their own games with zero DRM and still dominate the top selling charts.

But there's no way to download your games without Steam, right?

The only ways I'm aware of to download your Steam games are through the Steam desktop client or using the SteamCMD command-line client.

Edit: I think people really sleep on SteamCMD, I really like using it.

https://developer.valvesoftware.com/wiki/SteamCMD

From a developer point of view GOG has the most hoops to jump through of all PC platforms, for very few users, so it wouldn't make a lot of sense

Itch.io would be more reasonable. It doesn't have a lot of users, but it is trivial to setup, and doesn't require a launcher

I wish the Steam page had some pictures of the diorama setup from outside. It is a very original idea so play to its strenght and show more of what makes it cool.

The steam marketing part is interesting. You said your post got viral, can you share the hackernews submission post? Can't see in your profile.

Want to understand the user journey, as if people show interet they should check the game out.

I'm pretty sure it was this post, which I found by searching the domain: https://news.ycombinator.com/item?id=41884283

Dave Turner is making a point-and-click game with no CGI. All the locations are real-world dioramas he built and photographed. Special effects are done with stop-motion animation of felt.

Dave says: Often when you watch videos of people doing crafts on the internet, they're people who have been doing it for years. They don't make mistakes, or if they do, they don't really show them to you. I haven't been doing any of these crafts for years. So I've made lots of mistakes, and I want to show them to you, so you don't get the impression that this is easy.

Last thing we need on HN is low quality twitter bots reiterating what's been written.

The irony of course here is that _mjd_ is a fantastic author and contributor, and not a Twitter bot. That said, it does read like an AI summary.

Note that GP is the submitter of the article. I don't read this as low-effort engagement bait, but as an explanation of why they found this article worthy of sharing on HN.

This low-quality Twitter bot is very hurt by your cruel words.

> With seven rooms photographed so far, my repo right now is 53 GB, since I'm storing all of the photos that go into each focus-stack, as well as the raw (xcf) versions of each shot so that I can edit and color-correct without repeated JPEG compression.

I wonder how the author manages this bloat if he has to pivot to a different workstation. LFS enabled in his GitLab instance? Shallow clone? Just swallow whole??

you can copy/move/clone the local repo you already have. no need to clone from remote just to switch to a different machine.

To clarify, assume in good faith that pivot in context means compelled towards involuntary action---i.e. unexpected loss of local data, as opposed to a matter of routine shift in workflow---where the remote forge is being tapped for its secondary function as off-site backup.

The question was more intended to probe remote UX resilience near the limits of pain when faced with massive git repos which primarily capture binary assets/artifacts.

i find that interpretation a bit of a stretch. you could have expressed that more clearly. especially given that the obvious solution is to have local backups. what else could you do?

i have been discussing this on HN before in the context of a multi TB archive of photos and videos (part of which happened to be for a film project, so similar considerations apply). raid, external disk, and remote copy. for the film footage, everyone working with it has a full copy too, even if they don't need all of it. just in case. if you need to work with the data there is pretty much no other option.

Love the creativity here, this is novel game level design.

Can't wait to see the finished game. The blog posts are a part of the journey

Dave has a followup post in which he replies to some of the comments here.

https://novalis.org/blog/2025-03-26-responses-to-the-hacker-...

Love, your low-quality Twitter bot

Lol, the technique directing the art. Now everywhere a observer behind the wall would love to peep, there is a picture or lamp in the diorama. Basically , the artstyle subconsciously went towards subconciius thriller/horrorhouse with that .

Reply on news.ycombinator.com