×
  • remind me tomorrow
  • remind me next week
  • never remind me
Subscribe to the ANN Newsletter • Wake up every Sunday to a curated list of ANN's most interesting posts of the week. read more

Forum - View topic
INTEREST: 'Deep Anime' A.I. Creates Talking Animation Based on One Image, Voice Recording




Note: this is the discussion thread for this article

Anime News Network Forum Index -> Site-related -> Talkback
View previous topic :: View next topic  
Author Message
Kicksville



Joined: 20 Nov 2010
Posts: 1168
PostPosted: Mon Aug 12, 2019 6:00 pm Reply with quote
Well that's...barely impressive.
Back to top
View user's profile Send private message Visit poster's website
Ashen Phoenix



Joined: 21 Jun 2006
Posts: 2904
PostPosted: Mon Aug 12, 2019 9:44 pm Reply with quote
It looks kind of... terrible?
Back to top
View user's profile Send private message My Anime My Manga
Key
Moderator


Joined: 03 Nov 2003
Posts: 18135
Location: Indianapolis, IN (formerly Mimiho Valley)
PostPosted: Mon Aug 12, 2019 10:13 pm Reply with quote
This may not look impressive, but the industry is making strides. At a convention earlier this month I encountered an AI-driven "Waifu Generator," which could generate original pictures of your ideal waifu based on specs you enter in (hair color, preferred build, etc - and yes, a "Husbando Generator" was also in the works). The pics it generated were indistinguishable from something hand-drawn by a human and better than the bulk of the illustrations which typically pop up in Artist's Alleys.
Back to top
View user's profile Send private message Send e-mail Visit poster's website My Anime My Manga
harminia



Joined: 24 Aug 2015
Posts: 1987
Location: australia
PostPosted: Tue Aug 13, 2019 1:50 am Reply with quote
Key wrote:
This may not look impressive, but the industry is making strides. At a convention earlier this month I encountered an AI-driven "Waifu Generator," which could generate original pictures of your ideal waifu based on specs you enter in (hair color, preferred build, etc - and yes, a "Husbando Generator" was also in the works). The pics it generated were indistinguishable from something hand-drawn by a human and better than the bulk of the illustrations which typically pop up in Artist's Alleys.


I saw that. It did look cool and the technology behind it was quite well done but I had to wonder where the art came from. Was it acquired legally? I also didn't like that they presented it as a rival option to commissioning an artist.
Back to top
View user's profile Send private message
Netero



Joined: 10 Jun 2018
Posts: 159
PostPosted: Tue Aug 13, 2019 2:07 am Reply with quote
Given that software has existed for years that matches lip flap movement to a vocal track, this is really unimpressive. All it's doing is blinking the eyes and making the same repeated mouth movement when there is speech. This is just another bad example of "it's AI, give me money".
Back to top
View user's profile Send private message
Haterater



Joined: 30 Apr 2006
Posts: 1727
PostPosted: Tue Aug 13, 2019 4:54 pm Reply with quote
I think it'll get better in time. Not sure if I'll go for such a thing if its starts to replace people or do voices of those who passed away.
Back to top
View user's profile Send private message My Manga
ninjamitsuki



Joined: 15 Sep 2007
Posts: 589
Location: Anywhere (Thanks, technology)
PostPosted: Tue Aug 13, 2019 5:38 pm Reply with quote
Crazytalk did the exact same thing, and better...
Back to top
View user's profile Send private message
Ouran High School Dropout



Joined: 28 Jun 2015
Posts: 440
Location: Somewhere in Massachusetts, USA
PostPosted: Tue Aug 13, 2019 9:04 pm Reply with quote
Key wrote:
This may not look impressive, but the industry is making strides.

A touch of uncanny valley here, but I'd like to see another demo five years from now. Especially if a "first draft" can be stored and later refined by a human animator.
Back to top
View user's profile Send private message
I_Drive_DSM



Joined: 11 Feb 2008
Posts: 217
PostPosted: Wed Aug 14, 2019 10:48 am Reply with quote
It does seem rather unimpressive but I see a good application for something such as this in individuals creating basic games for themselves. Think a Ren Py visual novel where you possibly have character, background, and text assets created but cannot voice (or find people to voice) multiple characters. In that sort of thinking it would be relatively easy for someone to create a Doki Doki Lit Club style (as in length) visual novels by themselves that also has animated and voiced characters.

I also do not think it would necessarily take away from the VN industry but possibly supplement it. VNs flood the game market, and being able to push out more titles by smaller crews could see more ideas come to fruition, even if they're ultimately doujin efforts.
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Anime News Network Forum Index -> Site-related -> Talkback All times are GMT - 5 Hours
Page 1 of 1

 


Powered by phpBB © 2001, 2005 phpBB Group