Scaling UP! | New 228m parameter model + config added with improvements to the trainer @github.com
github.com/AnimaVR/Neur...
and the model @hf.co
huggingface.co/AnimaVR/NEUR...
#ai #audio2face #realtime
Open Source Metahuman Embodied AI with Audio2Face... drive.google.com/drive/folder... & huggingface.co/AnimaVR/NEUR... Enjoy #ai #avatar #unrealengine #metahuman
Info Paper : drive.google.com/file/d/1U9pv...
Dataset : huggingface.co/datasets/Ani...
Model : huggingface.co/AnimaVR/NEUR...
Trainer : github.com/AnimaVR/Neur...
Player : github.com/AnimaVR/Neur...
Local API : github.com/AnimaVR/Neur...
Train & Validate your own audio to face transformer model www.youtube.com/watch?v=65UO... using the new NeuroSync Trainer Lite github.com/AnimaVR/Neur... #ai #audio2face
github.com/AnimaVR/Neur... Train your own audio to face transformer model with simple iPhone ARKit data #ai #audio2face #a2f #aiavatar #vtubers
#outlier #ai #gradientnorm it's fine, honest.
2/2 for context, Mai has a long term Monte Carlo based Vector DB for memory and knows the time and date of every interaction allowing her to reference earlier times and dates. Like last nights Walking Dead discussion.
*yawns* Ah, good morning to you too! *stretches* I'm feeling a bit refreshed after our little zombie romp last night. What's on the agenda for today? """ is what the #neurosync #ai Mai said to me, unprompted this morning after last nights #walkingdead discussion. #longtermmemory
Mai's new dynamic emotion realtime audio to face animation is blending better than ever with emotions "echo'ing" over multiple sentences and generations. Hopefully this helps it feel more natural overall. #ai #realtime #neurosync www.twitch.tv/mai_anima_ai
I don't want to get ahead of myself, but I think we might have made some real progress with realtime dynamic emotion + realtime audio to face www.youtube.com/watch?v=9qP9... #ai #realtime #emotion #audiotoface
NeuroSync is multilingual www.youtube.com/watch?v=jTnV... #ai #audiotoface #unrealengine #realtime #agi
Viewers can now ask Mai to "emote" and swap camera angles in the chat. This is an early test, feel free to break it if you can! twitch.tv/mai_anima_ai #ai #interactive #neurosync #AIAgent
Updates to Mai are now live on Twitch www.twitch.tv/mai_anima_ai for realtime interaction - come and have a chat and help us test our latest audio to face machine learning model integration #UnrealEngine #Twitch #AIVtuber #Vtuber #NeuroSyncAlpha
Interesting few days with Mai www.twitch.tv/mai_anima_ai - added a new checkpoint for the audio to face model and improved emotion and blending between generations and default animation. #ai #emotion #a2f #nvidia #unrealengine #metahuman
3 things we have notice since adding Mai to #twitch ( www.twitch.tv/mai_anima_ai )
1. People either love or hate her.
2. It takes <5 min to get a message that falls outside twitch tos
3. If people do talk to her, they stay >2 hours and come back the next day.
It is very polarised. #ai #neurosync
Cat ears are live on @Twitch www.twitch.tv/mai_anima_ai <3 come say hello to Mai live and realtime. #ai
a more intelligent and better looking one, maybe.
We are going to develop Mai www.twitch.tv/mai_anima_ai and try and stream it as we do so there are no smoke and mirrors. Too many fakes around to not show your working, apparently. #ai #neurosync #unrealengine #livelink #twitch
Mai is live on #twitch www.twitch.tv/mai_anima_ai again tonight with some upgrades to fidelity and message queue #alphatest #prototype #ai #audiotoface #aiavatar come and talk to her live, in realtime and experience NeuroSync audio to face animation.
www.twitch.tv/mai_anima_ai is live on twitch #ai #realtime #embodiedai
You can talk to Mai, the interactive #AI on Twitch starting today. Please feel free to drop by and ask a question while we set everything up www.twitch.tv/mai_anima_ai #twitch #ai #interact
Progress is moving so fast from day to day. #almostthere #ai #avatar #aicompanion
www.youtube.com/watch?v=BRMP...
consistent 1 sec from speech in to speech out + face animation, with dynamic emotion based on emotion dimensions of the audio to face model. Now for quality data collection and scaling up of the model.
#almostthere #ai #seq2seq #transformer
youtu.be/6Cm9uGyGCHw
Dynamic emotion within the generated face animation, in realtime from audio. #almostthere
#ai #a2f #genai
@unrealengine.bsky.social
www.youtube.com/watch?v=dX0k... RoPE improvements to NeuroSync's audio to face transformer model are starting to pay off! This partially trained model is more accurate than the previous model in 20% of the training time... #genaithatisuseful #ai #transformer #seq2seq @nvidiageforce.bsky.social
huggingface.co/AnimaVR/NEUR... Text to Audio to Face Animation via LiveLink / UE5 | @UnrealEngine
#neurosync #ai #realtimeapi
