On his Rap God track, the effervescent witty Eminem points to a common critique of his delivery style: "They said he shaved like a robot, so call me Rap-bot; behind … "and so on.
It is a fairly common critique in hip-hop, applied to many less famous than Eminem. The idea of a "robotic" delivery style is very serious: your delivery is monotonous, the rhymes are obvious, you don't like the ear.
Welp, it looks like we may need to rethink things, thanks to an AI voice company called Replica and the semi-damn result of his employees being involved in what was apparently a "hackathon" (thank you, Virgin).
Interestingly, the rap itself comes from an individual, distributed among the different actors in AI. Replica employee Shreyas told PCG that "a member of the replica team performed all the rap bits. After passing the source recording to our system, we can apply the tone / cadence / performance in any of our more than 50 (and growing) AI voices ".
It also has some fun lines.
"So maybe you'll see that I'm a virtual actor,
And I'm paving new paths like a literal tractor. "
Not bad, although what would make it perfect would be an AI with the voice of Alan Partridge, the best DJ in Norfolk.
"The video itself is the result of something produced internally during a hackathon, using a feature that is currently being developed," explains Shreyas. "The feature itself is just about applying the same cadence / sync / delivery from a source recording to any of our AI voices. The fact that it can be pushed to remove a rap is a lot of fun and an unintended consequence of the feature that we initially only intended to add perfectly synchronized lip syncing to a video or 3D animation. For example, here's the same feature applied to Unreal Engine metahumans. "
(embed) https://www.youtube.com/watch?v=s5S3WMkdq7U (/ embed)
The feature will be released later this year, after debuting at GDC in July. The idea is basically to make it possible for developers to have a large cast of characters played by voice, with precise lip synchronization and intonation, without having to go through the process of animating and recording each line individually. Depending on the results, it could be a big breakthrough for smaller developers: now we’re laughing at virtual rappers, but in a few years this technology could be half the games we make.