MindVideo is a two module pipeline with each module trained separately and then fine tuned together. The model progressively learns from brain signals, deeply understanding the semantic space through multiple stages in the first model. This reports that their model outperformed other state of the art approaches in both semantic mapping and generation quality by 66% and 41%.
Researchers have developed a new method called Mind-Video for generating semantically accurate video from fMRI scans. Additionally, on todays Brief, NLW covers:
The most important announcements from Microsoft Build (including AI coming to Windows 11 and ChatGPT + Bing)
Google Bard images
Anthropic $450m Series C
Naval Ravikant's Airchat
The AI Breakdown helps you understand the most important news and discussions in AI.
Subscribe to The AI Breakdown newsletter: https://theaibreakdown.beehiiv.com/subscribe
Subscribe to The AI Breakdown on YouTube: https://www.youtube.com/@TheAIBreakdown
Join the community: bit.ly/aibreakdown
Learn more: http://breakdown.network/