When you set up the “meta human” do you feed it loads of images of the person, and from different angles and poses and expressions so it can learn?
Also, are you running this in a local model or online? If its online how do you feel sharing your wife’s likeness with the AI machine? 
Just asking because this is very cool and I want to play around with it But not sure I want my family’s pictures sitting with AI…
Looking at it theres no getting away from it, the banks already has your facial biometrics, apple already has it, cause that’s the primary way to unlock iphones etc. if the face is at the right angle then it needs minimal data cause it has enough to generated a close to accurate 3d model but obvious the higher the detail that you supply it the more accurate outputs you’re going to get.
Im planning a lara croft one now, but still working out, if im going to use a real reference, a reference from the game and have prompt created for the pose or if im manually going to specify how the pose must be
Everything is rendered online, i dont think current desktops have enough grunt to pull this off yet
That’s really cool, thanks for sharing.
Yeah, thanks!
I went down a bit of a rabbit hole with LORA and Dreambooths and things, and there might be a way to do it locally. We have an AI machine at work I can use to do the dreambooth training and then run my new model at home Will see how well that works before I share anything (if at all)
Only using prompts doesnt work for me the images always look sterile or synthetic, thats why i always use reference data, for increased realism
This was done without reference data only using prompts. and you can see it immediately thats its got a fakeness to it
My wife is a grade 3 teacher and yesterday I introduced her to Sora to help her create resources for teaching. She used buy these things from teacher sites. Within minutes she was churning our things like this:
Can the kids use AI to answer them? 
That’s very neat! Can you share the prompt that she used to generate those? I’d love to see how it’s structured.
create task cards 8 per page using sea creatures as the theme and giving subtraction problems in the number range 20 - 50. Label the cards with the letters from Q-X and the word Subtraction on each card
It will never happen but i can always dream
Did a render, using gemini pro and chatgpt to see what soul blade could look like if its properly remastered
Some of the recent ones
Street fighter 2 reimagined
Continued attempts at doing ultra realistic photos
What is Tia Carrere doing in Cape Town?
Shame, I tricked my daughter with my coffee art AI 
I dont blame her! The shadows etc is great!
That is really my cup, but the animation is created from a static photo.
