OK, made a little bit of progress with this today.

Out-of-bounds exceptions for lemming indices are now handled by displaying the expected vs. actual lemming index in the window caption, like this:


So, the replay has an action for lemming 11, but there are currently only lemmings 0-5 available in the level.

Also, when loading a replay, RetroLemmini falls back to matching by level name alone if it can't find an exact match for pack title, rating and level name (this has been in effect since 2.7). This does occasionally result in mismatched replays being played, so let's show the expected name + pack title in the caption. The purpose of this is to make it easier to see which level the replay is actually for:


The replay is for "Take A Running Jump" from Lemmings, But With Lemminas!. This pack can't be found, so RetroLemmini has loaded "Take A Running Jump" from regular Lemmings.

Whilst there's still no guarantee that the replay will play without issues, the above tweaks at least help matters by reducing unhandled exception popups, whilst providing better user feedback so that issues can be more easily identified and fixed.

Meanwhile, I'll keep looking into what can be done to increase backwards compatibility with older replays.

A question as old as custom lemmings levels, this one.

The difficulty of custom levels (as Proxima has already pointed out) is greater than that of the original games, but if we stretch the graph so that Fun still means "as easy as the game can be" and Mayhem still means "as difficult as the game can be", the following questions can provide a good barometer to sort levels out:

1) Does the level expect the player to find a single intended solution? Yes = 1, No = 0
   1b) Is the level a tutorial level, or a level where the solution is intended to be obvious? Yes = -1, No = 0
2) Does the skillset provide more skills than are required to solve the level as intended? Yes = 0, No = 1
3) Is the layout of the level specifically tailored towards patching out all other possible solutions than the intended one? Yes = 2, No = 0
4) Does the level's intended solution feature a specific trick that the player must know/discover in order to find the solution? Yes = 3, No = 0
5) Once the solution is known, how easy* is it to play the level in real time without using the pause button? Easy = 0, Not easy = 1
6) Are there any assignments which cannot be made without the use of the pause button, L/R arrows, or any other player assists such as framestepping or skill shadows? Yes = 2, No = 0

*For the purposes of this question, if the level provides little to no challenge when playing out the known solution in real time, it's "Easy". If executing the solution presents any significant challenge at all during real-time play, it can be scored as "Not easy".

If the level scores more than 0, it's probably at least Tricky difficulty. The higher the score, the more likely it is to be Taxing or Mayhem, keeping in mind that the highest possible score from the questions above is 10.

The questions matter in combination as well. For example, if questions 1, 2 and 3 all score positively, I'd say it's at least Taxing if not Mayhem. If questions 1 and 4 score positively, but 2 and 3 score 0, it could be a more difficult Taxing level or even a Tricky level depending on how lenient the skillset is.

Really, measuring difficulty is a learn-by-doing process. The more levels you make, and the more solutions to your own levels by other players that you see, the better an idea you'll have of how difficult your level is.

As a player, I always appreciate it when a pack starts off with a bunch of easy, open-ended levels just to get me going. It's a good way to get to know the author's level design style and get warmed up for the later levels. Like a professional piano player doing a few scales or playing a few easier pieces to warm up before a concert, I see it as a necessary part of the process.

Anyways, keep going, keep creating. That's the best way to learn.

Quote from: Airwave on February 27, 2026, 11:07:07 PMI'm new to the forum but I've been playing Lemmings on and off since the 90s. I got back into it in the last few month when I discovered RetroLemmini and have been playing a lot lately.

Welcome to the Forums! Glad you're enjoying RetroLemmini

I've replied to the replay concerns here.

Thanks for reporting

Quote from: Airwave on February 27, 2026, 11:07:07 PMWhile writing this I just discovered the replay with 17 saved, which I recorded in version 2.7, doesn't work correctly in 2.8 and becomes unsynchronised.

I'm unsure why this is happening. The level itself and the game physics are both identical between 2.7 and 2.8, so this desync bug is fairly concerning. We'll see if it happens with any other levels/replays.

Quote from: Airwave on February 27, 2026, 11:07:07 PMIt turns out the recording in 2.8 crashes while playing in 2.7.

This one's an out-of-bounds exception, these are unfortunately quite common in the Lemmini codebase. I'm working on better error reporting so that we see something like "there are fewer lemmings than expected" rather than a generic error, but it's a one-at-a-time kind of process.

Quote from: Airwave on February 27, 2026, 11:07:07 PMIt seems all other replays are fine when played in 2.7 and 2.8 so I can only assume that glitch causes problems with replays.

It's possible, but the expectation is that all replays should be cross-compatible between all RL versions after 2.6, which was the last time physics was updated.

Quote from: Airwave on February 27, 2026, 11:07:07 PMI noticed the replays in this thread from 2021 don't work in RetroLemmini.

I think I know why this is. The Flow Control replays from 2021 that you're referring to were created in SuperLemmini 0.104, which didn't have timed bombers. If you play back the replays with timed bombers enabled, the actions will desync. If you disable timed bombers and then load the same replays, they play back as expected. RetroLemmini has fixed this for newly-created replays (any replay made since RetroLemmini 1.0 accounts for the timed bomber option in the replay), but it may be worth implementing a failsafe if a 0.104 replay is detected - in this case, RetroLemmini should assume untimed bombers rather than defaulting to user option. <--- EDIT - this has now been implemented.

Incidentally, versions 2.7 and 2.8 both report on replay compatibility in the window caption. When you play back a replay, it may show something like this:



The above indicates that the current RL version is 2.8, the compatibility version is 2.6, and the replay version is 0.104. "Compatibility version" refers to the last physics update, so the general expectation is that replays made since that version should play flawlessly in the current version. The replay version is 0.104, which is well before 2.6, so there's a chance of errors occurring during playback.

I'll look into it and see if there's more that can be done to increase backwards-compatibility, or at least better error feedback.

Quote from: zanzindorf on February 28, 2026, 01:31:34 AMWhich one do you like better?

Thanks for making these! I probably prefer the straight up 32x32 one, but both could work.

Quote from: zanzindorf on February 28, 2026, 01:31:34 AMI don't know what you had in mind for art style ... I'm leaning towards simple, cartoony 3D renders?

Yes, ideal. By all means have some fun with it and see what you come up with. I'd say 32x32 is probably a good size to aim for if you're working with large 3D renders.

This looks like a fun challenge! I'd like to take a crack at it, though time might be spotty this weekend. I whipped something up in Blender/Krita for the replay icon.

Here's one with a dead-on camera angle.


Here's one with a slightly angled view, to hit the 32x24 aspect ratio a bit better.


It's a 3D model, so the lighting and camera angle can be whatever we like.


Which one do you like better?

I don't know what you had in mind for art style, but I have AnimationFactory.com on the brain atm, so I'm leaning towards simple, cartoony 3D renders?

Ah OK, good to know. Thanks for the explanation.

Quote from: Airwave on February 27, 2026, 11:07:07 PMHowever, I then checked YouTube and discovered there's a solution with 17 saved using a very odd glitch I've never seen before (if you turn a miner into a bomber and time it so that they would explode just before breaking through the last row of pixels, they die but there isn't an explosion and no hole is created).

That's not quite what happens. When the miner-bomber oh-nos, because of a positioning glitch (during part of the miner's animation, his position is below where it should be) he notices that there is no ground under him and falls. On this specific level, he falls out of the level before he can explode; if that didn't happen, he would explode like any other faller-bomber.

The same glitch can be used with a blocker instead (not on this level, of course) -- a miner can become a blocker on air and instantly transition to a faller. This was used in the first 100% solution to Wicked 6, although I think it has been achieved glitchless since then.

Hi  Guys,

I'm new to the forum but I've been playing Lemmings on and off since the 90s. I got back into it in the last few month when I discovered RetroLemmini and have been playing a lot lately.

I'm currently working my way through Oh No! More Lemmings as I never played it before (only had a SNES back in the day) and have loads of replays I'll share once I've completed Havoc.

In the meantime I'll share my replays of Havoc 10 - Flow control, which I just finished working on. I managed to save 16 with two methods (both with the same number of skills). However, I then checked YouTube and discovered there's a solution with 17 saved using a very odd glitch I've never seen before (if you turn a miner into a bomber and time it so that they would explode just before breaking through the last row of pixels, they die but there isn't an explosion and no hole is created). I managed to recreate this solution in RetroLemmini and saved a replay.

While writing this I just discovered the replay with 17 saved, which I recorded in version 2.7, doesn't work correctly in 2.8 and becomes unsynchronised. I'll record it again in 2.8 and upload both. It turns out the recording in 2.8 crashes while playing in 2.7. It seems all other replays are fine when played in 2.7 and 2.8 so I can only assume that glitch causes problems with replays.

One last thing, I noticed the replays in this thread from 2021 don't work in RetroLemmini. I guess they only work with older versions. I'm currently using version 2.8 and started saving replays with version 2.7.

Video of Havoc 10 - 17 saved with glitch: