Friday, May 31, 2019

Machine Learning Remasters Mega Man X4 and Mega Man 8 Cutscenes in Full 1080p


If you were a little bummed out about the quality of Mega Man X4 and Mega Man 8's animated cutscenes in Mega Man X Legacy Collection and Mega Man Legacy Collection 2, then you really need to see this.

A one VJ Ravenlight has used machine learning AI to "remaster" the opening cutscenes from Mega Man X4 and Mega Man 8 in full 1080p, extracted from their original 320x240 source. You can watch X4's HD intro above (Japanese version here) and Mega Man 8's (Rockman 8 ver. only for now) right here.

Mind you, these aren't *perfect* conversions. It's next to impossible to truly remaster these cutscenes without access to the original film or cels. Nevertheless, it's true: these cutscenes have never looked this good in any official capacity.

VJ Ravenlight has informed me that he's currently working on 1080p versions of every animated cutscene from EVERY PlayStation-era Mega Man game (MM8, X4X6). These videos will be posted to his YouTube channel in the near future.

All in all, this is some real admirable work. I can't wait to see what comes next!

12 comments:

  1. "It's true; these scenes have never looked this good in any official capacity. Bravo, AI."
    It looks worse, by a very large margin. These are not good examples of what machine learning is capable of

    ReplyDelete
    Replies
    1. Then, can you provide a better version of it? Since you said this is pretty bad. I would like to see it.

      Delete
    2. Worse? Really? I, uh, don't know about that.

      Delete
  2. Nice work, but if he really wants to impress me... he should try to do something with the PS1 version of X3.

    ReplyDelete
  3. This is super neat! Thanks for sharing the news.

    ReplyDelete
  4. Awesome stuff! They really need to come out with a mmx anime.

    ReplyDelete
    Replies
    1. I've been wanting a mmx anime to happen since the 90s.

      Delete
  5. Yeah the quality was never really looked perfect but this intro...amazing for ever!
    In despite of this, great job.

    ReplyDelete
  6. That's some pretty impressive coding. I've never really delved into machine learning, but this product is outstanding.

    ReplyDelete
  7. It's nice to see machine learning be used this way. Cool stuff!

    ReplyDelete
  8. I would love to know what kind of training data set was used. I wonder if there is a paper or report with the details.

    ReplyDelete
  9. Not too shabby!

    ReplyDelete

Keep it friendly. Disparaging, belittling and derogatory comments are not permitted.