Google AI - Parti: parti.research.google
"The 20B model especially excels at prompts that are abstract, require world knowledge, specific perspectives, or writing & symbol rendering."
+ Pathways Autoregressive Text-to-Image model
+ ViT-VQGAN, Lingvo, GSPMD, PartiPrompts (P2)
74
Schmitt: ‘Blueprints for Intelligence’ (2022):
philippschmitt.com/blueprint…
"Instead of projecting onto a metal humanoid, a poetic reading of metaphor & symbolism in AI diagrams prompts us to consider how their creators think about cognition."
#NeuralNets #Perceptron #LeNet5 #AlexNet
1
VentureBeat: "2022 - The Year of Generative AI":
venturebeat.com/ai/how-2022-…
"Contrastive Language-Image Pre-training (CLIP) became pivotal in text-to-image generators. CLIP + diffusion were used in OpenAI’s DALLE-2 to generate high-resolution images with stunning detail & quality."
1
Imagen Video: imagen.research.google/video…
"Generates high resolution videos with Cascaded Diffusion Models: 1280×768 pixels @ 24 frames per second."
+ T5 Text Encoder
+ Video Diffusion Model
+ Temporal Super-Resolution (TSR)
+ Spatial Super-Resolution (SSR)
+ Video U-Net Architecture
Generating Automatic Video Subtitles from any Language with Whisper AutoCaption: blog.paperspace.com/automati…
"The holy grail of speech integrated NLP is the ability to directly transcribe audio into words in real time with a high degree of precision."
#Whisper #OpenAI #NLP #MoviePy
1
D-Wave Quantum Proof of Concept (QPoC):
aws.amazon.com/marketplace/p…
"explore selected optimization use cases through detailed problem discovery, problem definition, development of an initial prototype solution, & testing with representative sample data."
$QBTS dwavesys.com/
NightCafe Studio: creator.nightcafe.studio/tex…
+ VQGAN+CLIP Text to Art Generator
+ Stable Diffusion AI Image Generator
+ CLIP-Guided Diffusion (Coherent algorithm)
+ New: DALL-E 2
+ "free to generate up to 5 artworks per day"
+ React, TypeScript, NextJS, NodeJS, & Firebase
#NightCafe
A Brief History of Neural Nets and Deep Learning (2020):
skynettoday.com/overviews/ne…
EbSynth: ebsynth.com/
Photogrammetry in Hubs (2022):
hubs.mozilla.com/labs/photog…
-> Meshroom, DarkTable, Polycam, & Blender
-> "the immersive web"
#NeRFs #ImageNet #MozillaHubs #AliceVision
MetaHuman: unrealengine.com/en-US/metah…
"complete framework that gives creators the power to use highly realistic human characters in any way imaginable... includes MetaHuman Creator, a free cloud-based app that enables you to create fully rigged photorealistic digital humans"
#unreal
1
The Illustrated Stable Diffusion:
jalammar.github.io/illustrat…
"The ability to create striking visuals from text descriptions has a magical quality to it & points clearly to a shift in how humans create art. The release of Stable Diffusion is a clear milestone in this development..."
CogVideo - Text-to-Video Generation:
github.com/THUDM/CogVideo
arxiv.org/abs/2205.15868
Imagen Video: imagen.research.google/video…
Large Pretrained Language Model × Cascaded Diffusion Model:
imagen.research.google/
Phenaki (2022) - Generating Videos from Text:
phenaki.github.io/
1
1
Imagen Video: techcrunch.com/2022/10/05/go…
"builds on Google’s Imagen, an image-generating system comparable to OpenAI’s DALL-E 2 & Stable Diffusion."
+ trained on 14 million video-text pairs + 60 million image-text pairs
+ LAION-400M image-text dataset = aesthetics
#Midjourney #COCO
1
1
DreamBooth:
dreambooth.github.io/
Stable Diffusion v1.4:
huggingface.co/CompVis/stabl…
Deforum Stable Diffusion:
colab.research.google.com/gi…
Dreambooth on Stable Diffusion:
github.com/XavierXiao/Dreamb…
Fast Stable Diffusion + DreamBooth:
github.com/TheLastBen/fast-s…
#DALL·E2 #Imagen #LAION #CLIP













