The creation of this TAS of Rayman 3 (GBA) started on the 9th of August 2021 using the latest version released of the emulator BizHawk which was 2.6.2.
I used the default setup of the GBA emulation from the downloaded file BizHawk-2.6.2-win-x64.
The ROM of the game I used is entitled 0985 - Rayman 3 - Hoodlum Havoc (U)(RDG).gba
I created the movie "Rayman 3 (USA) (En,Fr,Es) final TAS Any percent GBA.bk2" using the tool TAStudio.
About this Any% run:
English is the fastest language. The run starts at frame 866 (selecting an empty save file) and ends at frame 200319 (last hit on the final Boss).
So this run has a length of 200319 - 865 = 199454 frames.
The frame rate is about 59.7275 fps according to this submission, so I estimate the final time to be 55:39.40 (manually timed it 55:39.4 with LiveSplit anyway).
Both final dialogues and credits are skippable, but I kept them to show the end of the game. The A presses after frame 200319 do not count in the final time of this Any% run.
I've optimized this TAS the best way I could do. First, I used the best human strats known. Then I introduced many new frame perfect tricks such as:
- Single/double/triple frame jumps.
- Double hits on a single frame against Hoodlums and Pirates.
- Fast climbing.
- Fast walking and jumping recovery after laying on the floor.
And TAS only strats discovered such as:
- Tight shortcuts impossible for humans like in Void of Bones and Wretched Ruins.
- Fast damage boosting against Jano.
- Skipping all loops consistently in Prickly Passage.
- Zipping in The Mettleworks.
However, this TAS isn't optimal yet. Indeed, the RNG in the final level aren't the best. I found a seed that is 45 frames faster than the actual seed. Unfortunately I couldn't fell back on that seed despite many manipulations with free inputs. Else, I tried once to deoptimize a segment by 10 frames to get a better seed that was only 11 frames faster. Loosing my satisfaction in order to save 1 frame wasn't worth it. So I kept my fully optimized segment and slower seed.
Then, there are still two levels that aren't fully optimized: Magma Mayhem 1 and Magma Mayhem 2. I did my best to save frames by testing many routes. The actual result is the best route I could empirically find. Those two levels are designed like racetracks from Mario Kart, so very difficult to be optimized at full potential.
Finally, I didn't introduce many free inputs to entertain the audience during waiting segments because free inputs can lead to additional lag frames and change RNG seeds. I just did some stuff in Scaleman's Keep because I manipulated my best seed here.
I hope you'll enjoy this TAS.
: Claiming for judging.
: Removed blank input from the end of the movie.
: Great run, more than 7 minutes faster than RTA even within the same routes. Accepting.