The death load remover has been uploaded to GitHub! So far Anthony, PDub, and myself have had success with the update. Feel free to give it a shot and report if you have any issues.

https://github.com/blegas78/autoSplitters

I'm happy to share a preliminary test video showcasing load removal of a death! Here, the strong image features in the word "NEXT" are scanned in the tutorial screen. Once it appears, the IGT is adjusted to handle the prior black screen by subtracting the ~3.5 seconds, then remains paused until the screen turns black then shows anything else. This does not affect LiveSplit's RTA timer.

Looks like things have been going super smoothly for those that have been using the load remover! Last night was the first sighting of a false negative in PDub's stream, where the load remover resumed the timer prematurely due to terrible moth placement.

I have since downloaded clips form his stream, used the videos to reproduce the error, then tweaked the load remover to be even more robust to moths. I have uploaded the more robust configuration files (both French and English) to GitHub. I recommend everyone using the load remover to grab the latest files.

https://github.com/blegas78/autoSplitters

This kind of data is wonderful to have! If anyone else has issues with the load remover pausing/resuming at strange times then share the video with me so I can make the load remover stronger :)

Kevin700p tried the load remover with no success. After working together, it was discovered that they had French set for their language, resulting in a misaligned Back button icon. After switching to English, it worked perfectly. I also spent 5 minutes to make new .vas game profile specifically for French, so now both French and English are supported.

Thanks Kevin700p! This feedback is greatly appreciated so I can ensure better tool quality. If anyone else has load remover problems, feel free to reach out to me!

The LRT appears to work flawlessly for how it has been designed. Regarding the above findings, the LRT resulted in a time longer than IGT meaning that the timer should have paused more than it did. This means design clearly does not account for all loading instances. These instances are outside of the definition of the player opening the pause menu then invoking a cinema skip or restart checkpoint/encounter.

There are a few instances where the game cuts to black, not invoked by the player. I'm not sure how many there are in total, but here are two examples from DemonicRobots' run:

There may be other instances, but those black screens where 6 seconds and 7 seconds long, respectively. Assuming condition #3 in the prior post, this alone would reduce the potential LRT vs IGT error range to +00:01:03 to +00:00:04. We still don't know exactly when the IGT officially starts nor ends so those issues could still be present.

Are there any other instances of the black screens not invoked by the player? Should these be added to the definition of a loading screen?

After some recent discussion in the discord it was determined that the above assumption for when a load starts does not match ND's IGT. Instead, the assumption is that ND's IGT pauses whenever the pause menu appears, but ends as I assumed above.

This change has been made and DemonicRobots seems to have tested it with great success! His IGT time said 4:51 while the LRT showed a time of 4:52:16.

Now why is it off at all? Well we probably don't know exactly when ND's timer first starts nor when it ends. Thus how DemonicRobots started/ended the timer could be different. One thing I don't know is how ND rounds the seconds in their time. We can examine how far off the LRT was for that run by either 1 of the 3 amounts:

1. Seconds are rounded up: the IGT is 4:50:00-4:50:59 --- LRT off by +00:02:16 to +00:01:17
2. Seconds are rounded normally: the IGT is 4:50:30-4:51:29 --- LRT off by +00:01:46 to +00:00:47
3. Seconds are rounded down: the IGT is 4:51:00-4:51:59 --- LRT off by +00:01:16 to +00:00:17

I speculate that ND rounds their times down since they say "Less than 1 min" then transition to "1 min" I believe, but could be wrong. There could still be some times when ND pauses their timer that we simply don't know about, not captured by the rules set by the LRT.

Hello all! I'm the one working on the Load Remover Tool for LiveSplit for TLOU2. I'm hoping that this tool can greatly aid in the live-timing of runs.

I know that there are efforts to add seconds to the IGT, but even so that will only be checkable at the end of a run which significantly reduces the quality of running this game. Also, the IGT may pause the timer when the user is in a cut scene/menu, potentially not catering to the spirit of speedrunning. What ND considers to be IGT will be different than how submitted runs may be timed.

Thus, I would like to know how runs will be potentially timed for load removal. What is the exact definition of a load? I can imagine that a load is defined as starting on the frame just after "Skip Cutscene" or "Restart Checkpoint" is performed, up until the screen cuts away from black (excluding the moths). So far I have the video remover working very well in these scenarios. There may be other scenarios that also define a load and would like some some input.

The reason I bring up load definition this early is to potentially provide advice on what should be considered to be a load. If loads are defined how I described above, then the video load-remover will be trivial for me to develop. If it is not defined in this way then hopefully I can be in the loop in some capacity to provide information what would be easy to time.

I will be interested to hear what you all have to say! I'm looking forward to attempt providing useful tools to the community.