===============================================
Here's some Notes on Getting answers from different AI's
The You.com (first one) and Perplexity (Last one) are the best.

Link to Long answer from You.com :
https://you.com/search?q=Compare+the+architecture+and+algorithims+of+VASA+and+SadTalker+by+reading+the+respective+papers.+...&cid=c1_872a27ca-2dc6-4e77-8cfe-b9a21f76ae27&tbm=youchat

Same Question to Gemini 1.5 Pro
https://you.com/search?q=Compare+the+architecture+and+algorithims+of+VASA+and+SadTalker+by+reading+the+respective+papers.+...&cid=c1_65c1d660-27dc-4ad5-b98d-f639fc76a42e&tbm=youchat

Same Question to Llama-3-70b-T
https://poe.com/chat/27fdtqe96f5h0kv23yd Multiple Halucinations - lower quality than above

Same Question to meta-llama/Meta-Llama-3-70B-Instruct (Web Search Enabled)
https://huggingface.co/chat/models/meta-llama/Meta-Llama-3-70B-Instruct/
Link to Chat https://hf.co/chat/r/yrHJLw8
what's this? convolutional recurrent neural network (CRNN)

Same Question to GPT-4 https://copilot.microsoft.com/
answer not that good. Got the wrong VASA (not AI) saved to Copilot GPT4 Answer (VASA and SadTalker Comparison).txt
tried to include Papers, but they are too large

Same Question to Perplexity (Academic Mode + Pro Mode + Attached 2 Files on Papers) https://www.perplexity.ai/
Changed this line ==> Compare the architecture and algorithims of VASA and SadTalker by reading the respective papers I've attached.
(and took out links to papers)
==== Link to Chat https://www.perplexity.ai/search/Compare-the-architecture-B3Yn6KrgTc.jATiPzW2JCA
=====================Look into these:
connectionist temporal classification (CTC) loss
sequence-to-sequence loss.
mel-spectrograms
such as lip-syncing (Wav2Lip), facial expression (ExpNet), and head pose (PoseVAE).

Interesting Question: I asked You.com AI the following question in Research Mode and got a very thorough Answer:
Compare the architecture and algorithims of VASA and SadTalker by reading the respective papers.
Vasa Paper: https://arxiv.org/pdf/2404.10667.pdf
SadTalker Paper: https://arxiv.org/pdf/2211.12194.pdf
Summarize each of the architecture and algorithims used in the papers and then answer the following question:

Is it possible to iterate with transformers and get to the quality and a capacity of VASA with what’s available on hugging face? Especially playing around with this GitHub repo https://github.com/OpenTalker/SadTalker

Link to Long answer:
https://you.com/search?q=Compare+the+architecture+and+algorithims+of+VASA+and+SadTalker+by+reading+the+respective+papers.+...&cid=c1_872a27ca-2dc6-4e77-8cfe-b9a21f76ae27&tbm=youchat

Youtube Video LiJo https://www.youtube.com/watch?v=c23PRUHpIHE
8 Cognitive Functions: Around the House (funny)

@ScottzPlaylists (Comment)
If someone put together a DB of short clips demonstrating all the 8 cognitive functions and MBTI types,
that would be a great training dataset for AI to train on detecting that in other videos.
Then AI can build a larger database from watching videos and extracting the clips. (especially with Human Feedback)
Then we are on our way toward AI Detecting and Generating Personalities (MBTI, Big Five), and Cognitive Functions.
I would love to these detection's in YouTube Videos, Movies, and any social media videos

Good Youtube Videos to start with....
==https://www.youtube.com/@FrankJames
==https://www.youtube.com/@InternetLiJo

What's the closest Code to this? Will MS release the code it?

So Glad to find this...
Also your "Big Five Personality Traits" collection is nice.. I've been interested in MBTI for a long time.