I just thought, now that Rebecca has a couple weeks of adventuring her look needed a little update. Not so much the exuberant grin of before, still a relentlessly good natured sort. But a little more pragmatic now. She has the “NWN – Female, Rogue” voiceset, which has matched the portraits perfectly.
I like the new look. Definitely looks like the same character having matured a little. Still has the smile, but not quite as much childish innocence. Makes me want to do a study on aging with some of my characters. What would Blur look like as she ages? What did Roland look like as a teenager? Etc.
I like that voiceset as well. Neither Roland nor Kat had a good voice in the baseline samples, but mining NWN, BG1, etc produced some much more palatable options. If only the NWN didn’t *constantly* repeat that “bet you can’t see me now” phrase every time stealth gets rechecked….
Her first portrait sure made me laugh! But I thought by the time she hit 8th level mage she should look a little less geeky “I’m ready to do something now”!
and I generated so many of her trying to get a darn realistic looking crossbow…
So I end up using two from my first “I’m giving up on that now…” group.
LikeLike
Funny and frustrating! I don’t know if the current AI will learn? Or if new data will need to be added? Presumably we would need some way of providing feedback for that to work? (Ugh, not that you stupid computer!) But I have no doubt it will get better.
LikeLike
The current Bing AI is based on this one: https://openai.com/index/dall-e-3/
Its the third generation of DALL-E, and I’m sure there will be a fourth at some point. They’re always adding to the training set, tweaking the algorithms, etc. The big limitation on some of these is copyright/IP. Something there’s big holes because they’re simply not allowed to use some things in the training set. Others because they want to protect against it generating copyright-infringing content – like Mickey Mouse the dragon slayer. It’ll be wild to see where it goes in time.
LikeLike
I’ll assume a crossbow is out of its patent.
But yeah, the characters and licensing are obviously a big deal. Like I was saying the other day, I’d love to do “Indiana Jones as a Ranger” sort of thing. But I understand that will be problematic, at least from the more “reputable” sources for a while.
I wouldn’t be shocked if we reached the point where several companies and studios would have their “own” AI for that sort of thing (do your own Tony Stark as Tinkerbell only with the Disney generative AI!).
OOF. Now I need to get that image out of my head!
LikeLike
Bah. Got the hyperlink wrong somehow. Just change the “2” to a “3”
LikeLike
AI has done great things for DnD fans, both on the tabletop and off of it.
If my campaign of a few years back was still in full swing, I would have made ample use of these tools. From voices, to backgrounds, to character portraits, even story ideas!
LikeLike
Absolutely! I’m still playing around with its limits. The ability to do non-human, fantastical beasties is pretty amazing.
It still has some major limitations too! Like its inability to draw crossbows. Even some badly depicted swords on occasion. The biggest thing to me, would be to be able to save an image and tweak from there (obviously, that’s more specifically a Co-Pilot limitation than AI in general). I am looking at other AI options too.
We are in the early days of this. But already, the ability to bring to life things from the imagination (or my own PnP game!) is just amazing.
LikeLike
DavesGaming 
Leave a comment