The author of this run uses a plugin that allows him to go faster, so I tried it myself and it really noticeably affects the speed. If someone in the community is left unsubscribe about it want to know your opinion about it.

Will it be considered a cheat or not?

Anyone write what we thinking about it.

Since @MuKJIyXaRRR addressed me personally in one of his recent speedrun, once again reproaching me for abusing some "plugins" - quote: "Если ты плагины использовал, то я тоже использую" ("If you've used plugins, then I'm using them too"), and the title of the thread primarily refers to me - I decided to conduct comparative tests of different image rendering modes, especially since he kindly indicated his settings.

It seems that the initiator of this tread was confused by two things in my speedrun "Rookie 47:20":

1. The word "plugin" in the clarification "Video render plugin: Quad / hq3x";
2. A difference of one frame with the PB-video in the corner of the screen, when the image appeared at the start of the first race (after all, before that the menu went out synchronously); and it just so happened that it was recorded with anti-aliasing hq3x, and the run itself was in pixel Quad.

But first - a few introductory:

• If someone interested in establishing the truth for some reason is not satisfied with the quality of the videos I provided, or the data seemed insufficiently comprehensive, informative and visual, he can conduct his own tests and publish them in this thread.
• The position of the objects on the first race is formed when choosing the difficulty level or entering the password and does not change after, even if you go through the store and menu items for a long time. This is convenient for testing different settings on one seed of the track using a single savestate. And for this reason, it is so important to start any speedrun with a hard reset of the console, because there is an opportunity to "sneak a peek".
• All tests were conducted on VS. Mode, Rookie.
• After loading the save, I held down the accelerator button and held only it alone until crossing the finish line.
• The YouTube player allows you to step through the video frame by frame using the “<” and “>” keys.

First: what did I mean by the term “plugin” back then in the descriptions of my runs. This is a direct extract from the settings of the Kega Fusion 3.64 emulator; regular (anti-aliasing and not so much) graphic filters, which are defined in the “Render plugin” option in the “Video” tab; interpolation filter files have the .rpi extension (in case this is important). They are not included in the “delivery kit” by default, but their presence is provided; some emulators include such filters initially. I downloaded a Russified version of Fusion from a now defunct site, which already had filters; I did all Sega speedruns only on this build. Since the rules do not limit anti-aliasing or prohibit certain filters, it would be strange to reproach for this, and this is not the point at all. At that time, I was running Beyond Oasis and Comix Zone with identical graphics settings, and the communities of those games were moderated by smart people, including those who were well acquainted with the Mednafen emulator - zero complaints and questions for me.

Video 1. Kega Fusion. Compare the effect of the graphic filters that I enabled then (Quad and hq3x), and two variations of the Normal mode - with and without vertical sync. As a result, we get almost mirror images - only the one without VSync lags by one frame by the end of the race.

Video 2. Everything is the same, plus a number of aspects that are almost always active for me are additionally enabled. The sound is played in all windows. The images are synchronous; abnormally "go faster" is not observed again.

Second: where did one frame "go"? Along with an elementary error in coding, including due to the possible disabled VSync, in RRR there is a different loading speed of dissimilar tracks. A trifle. But exactly one frame difference. Maybe more.

Video 3. Four races with different seed were generated. Sound is on in windows 1 and 2, which at one point starts to diverge due to Larry's changing lines. With the same graphics settings, windows 2 and 3 appear one frame earlier. Or... others... later? Is the glass half empty or half full?

Video 4. Finishing up with Kega Fusion. There are a number of other settings in the "Options" tab, the purpose of which is not entirely clear to me. Before this, everything except the indicator was turned off.

• Window 1 plays my PB "47:20", without sound.
• The image in window 2 "jumps" sometimes earlier, sometimes later. I think "Perfect Sync" behaves unstable in practice and should be kept off.
• At 0:32 seconds in window 3 (lower left) Larry says a line that is not in the other two.

Moving on to the Mednafen emulator stats. This version has been on my disk for many years: Doomday Edition 7A; 0.9.45.1.0.0.7a. Shell (GUI): Mednaffe NST 0.8.6.3. Doesn't match Miklouha's build, but it's hard to imagine that the fundamental functions of the Genesis core emulation are any different.

A few more introductory notes regarding the general configuration.

• In the "Resolution" section, the window mode scaling factor is set to 3.00 in X and Y, which results in video capture at 872x672 pixels. Therefore, YouTube projects are rendered at 1744x1344. The Kega Fusion window was recorded at 960x720, which gave more familiar proportions: 1920x1440.
• CPU overclocking multiplier is set to 1 (x1, no overclocking).
• I have a localized version, and the language in the emulator does not change. Although, I suppose, the translation of menu items is not essential here.
• Fusion and Mednafen support different savestate files, so it seems problematic to conduct verification on one seed.

Video 5. In the "Filters" tab, we alternately change the values to those specified by Miklouha. Top to bottom:

• (Bi)linear interpolation (0 everywhere)
• Stretch screen fill (0 / aspect)
• Special video scaler (hq4x / nny4x)
• OpenGL shader (0 everywhere)

The resulting image is not very synchronous; with an uneven, but stable difference of 1-3 frames until the end. Not such a mirror flow as Fusion. At 0:32, Larry can be heard in windows 2 and 3.

Video 6. To determine how strong the filter factor is, we compare three versions of the basic settings (0 / none). It has become better: only one frame of desynchronization is noticeable. We additionally compare it with the Normal mode of Fusion: there is a slight discrepancy in the soundtrack. The videos themselves at first “move away” and then synchronize, but after the fourth corner, due to the different physics of contact with the fence, the comparison of the two emulators can be considered incorrect.

Video 7. Finally, we look at what kind of OpenGL shader “sabr” is and how it “gets along” with filters, while comparing it with “none”. And... the same 3 frames of difference, which at a distance of “Rookie”, theoretically, can knock off 2-3 seconds. I also wanted to compare Fusion and Mednafen in another game, but, apparently, this is not necessary.

Now my opinion on all this.

• The statement "The author of this run uses a plugin that allows him to go faster" is slander. (Cheating in empty categories...) It feels like Miklouha has never opened Fusion, since he compared its "plugins" with Mednafen's filters, without providing his results. There is a difference, and apparently it may not only be a matter of “plugins”. There is no OpenGL mode setting in Fusion, not even in the .ini file. Well, or I'm blind...
• Fusion filters are just as "cheat" as the objective error of +/- two frames, caused not only by system performance, but also by the game itself. I will continue to play (or never again) as I am comfortable, unless there is a specific prohibit.
• I could consider that Mednafen runs should be performed with basic graphics settings, since, as it turned out, its filters process the image an 1-2 frames per minute faster/slower, but for this you need to conduct dozens of tests (and not one and a half minute ones, but full ones) to be sure of the advantage for sure. Probably, it is easier to implement this through TAS... Or, maybe, this is only on my system, and for others everything will be fine, and this "article" did not make sense.
• The video recording of the sent run should begin with a Hard reset.
• Comments from competent people who understand graphics rendering and the accuracy of emulation would not hurt on all of the above. After all, for some reason, emulators familiar to us are banned in certain communities...

Well, at least someone wrote in, thanks, man.