Black Mixture
patreon
CREATE Mind Blowing CHARACTER Animations in SECONDS!│Runway Act-One Tutorial
🕑 Added 2024-11-08 13:46:50 +0000 UTCWe recently tested out Runway's latest addition to Gen 3 Alpha called Act-One. The promise of being able to lip-sync videos based on an input image and input video is HUGE as it introduces a feature many animators have been dreaming of. A silver bullet solution for animating talking heads has been in development for decades, yet there hasn't been an approach as AI heavy and well implemented until now. There's still much to be desired, but Runway has made an interesting leap forward for the animation community with this release. So here's our quick review scores and findings of this update.
2D:
85/100
- Worst quality of the various input types, yet still very impressive as traditional methods of 2D animation would easily take hours to produce a few seconds of usable footage. Runway's Act-One can handle 2D images well, but there is still some lack of creativity when it comes to translating 3D motion to 2D spaces, particularly for the shadows and mouth movements.
3D:
87/100
- Medium quality of the various inputs. Possibly one of my favorites as it handles just about every aspect of the facial animation close to perfectly. If there wasn't a slight noticeable degradation of quality around the animated mouth and the rest of the character, this would be perfect.
Live Action:
90/100
- Best quality of the various inputs. There isn't anything to really complain about here. The model handled various characters of different genders and races with such ease and grace that I would have thought the footage was filmed by actors on a movie set if it were not for the static environments. The live action examples show just how close we are to achieving a near perfect result in what I can only imagine would be a few more months. Bravo Runway. Bravo.
Workflow used to generate the animated style images here in our previous supporter-only post.
