Welcome to the 186 new folks joining us since yesterday!

Today's email is a little longer than usual, it happens every Tuesday. Just remember, this is like a huge spread during the holidays. Grab your favourites first, dip in and out all day then leave the stuff you cannot possibly stomach. There are no prizes for looking at absolutely everything, don't panic 😉.

Let's get to it!

p.s. weird that people always read the p.s's?!

p.p.s. I set up a Discord³⁷¹ for anyone who wants to hang out, chat AI, share interesting things for this email etc. Come say hi³⁷¹ (if you can stomach another Discord).

p.p.p.s. got ya, you read this too right?

Prompt: List the topics in today’s email:

🤡 Create virtual characters💃 Move like Mona Lisa🫂 Shuttershock's CEO on embracing AI🌁 3D flythroughs from a single image

🫦 BEN'S BITES

The new Lexica Model looks INSANE⁸²⁸. It’s genuinely unbelievable that these were created with AI.

Training Stable Diffusion with Dreambooth⁶⁸¹ using 🧨 Diffusers.

Animating focal lengths²⁰¹ is now enabled in nerfstudio.

Adobe just rolled out a beta of AI-powered podcast audio recording and editing^314†, all on the web. It includes auto-enhanced audio, easy-to-invite guests with a link, and more to come.

Anton Teaches Packy AI - A ~weekly YouTube series²⁷⁷ in which Chroma's Anton Troynikov explains a key AI research paper to Packy - “Asking dumb questions to smart people is the role I was born to play”. If you want a deeper AI understanding than "WOAH," this is for you.

👋 Too many links?! I created a database for all links mentioned in these emails. Refer 1 friend using this link and I'll send over the link database.

Create, explore & shop A.I. generated products³⁴².

Daniel Ecker released another MEGA THREAD on AI.⁴⁶⁴

PhotoRoom raised $19 million for its AI photo studio²⁶⁸. It has a new generative AI feature, magic studio, to create stunning marketing images from any product photo.

Replicate is hosting a Demo Day¹⁸¹, alongside Fabian Stelzer. Demo’ing GPT3-powered prompt generators on Google Sheets, satirical pitch decks, and more.

text-to-Figma: build and edit Figma designs with natural language. Sign up for the waitlist here³²⁸.

Circle Labs³⁹⁹ just raised $4.2M from Lightspeed. It’s a platform that enables anyone to become a creator of unique, virtual characters powered by AI. The characters can talk to people on various platforms such as Discord and Twitter. Read their announcement here³⁹⁹ - and yep, sign up for the waitlist here¹²³.

Auto-generated Piano covers - Pop2Piano¹⁶⁴. You can use HF to enter audio²²¹ and get your own cover.

ThisDogExists⁴¹⁰ is a project to see if you can identify the real dog in a lineup. The real dog in the lineup is looking for a forever home.

Ever wanted to make Mona Lisa move? Well, Researchers at Cornell University have produced an animated image from a significantly different source image²⁸⁹.

Connect your database, ask a question, and get an answer. (waitlist^405†)

An interview with the CEO of Shuttershock on why they decided to embrace AI²²⁴.

DeepMind’s looking for applicants for its scholarship programme²²⁴.

The Stability Photoshop plugin²⁸⁷ allows you to generate and edit images using both Stable Diffusion and DALL•E 2 right inside Photoshop.

Meta-FAIR's NLLB-200 content translation system is getting used more and more by Wikipedia editors⁹⁷ and produces higher-quality results than other translation systems.

A little more info on the AI lawsuit²⁰⁹ against OpenAI, GitHub and Microsoft.

AI Avatars may have looked a little gimmicky and there were plenty of projects that effectively did the same thing…however, Pieter Levels’ AvatarAI has made $100k+ since launching 10 days ago³¹⁵. Impressive!

Transcribe music from YouTube videos³⁰⁸.

Subtitle any video to English¹⁸⁶, split, merge & adjust timecodes and get a final subtitle file.

New GoogleAI research with Berkeley produces high-resolution 3D flythroughs from a single 2D image²³⁶. The model is trained only on still photos and fills in missing pieces (like other sides of mountains), rendering with high realism.

Supervising AI systems¹⁹⁹ that potentially outperform humans on various tasks.

A new and improved Hugging Face Deep Reinforcement Learning class batch #2^150†! Starts on 5th December.

How to learn a visual walking policy²³⁴ that only uses a monocular RGB camera and proprioception to walk. No, not you, it’s for the robots.

A new method for video-and-language pre-training¹⁶⁷ that incorporates structural knowledge into the representation learning process.

A list of AI startups in San Francisco²⁸⁵.

Carbon Re has raised a $4.8m seed round¹⁰¹ for the world's first AI-powered platform designed to enable zero carbon manufacturing.

Dreambooth Extension for Stable-Diffusion-WebUI¹⁹⁷.

There are trade-offs between scale, multilingualism, and compression when training language models. This paper presents an experimental framework to study the impact of sparsifying multilingual models during fine-tuning¹⁰⁶.

The importance of the attention mechanism in the Transformer architecture¹³¹ and how it can be replaced with a simpler alternative.

How Hugging Face and ServiceNow tackle code-generating LLM challenges²⁶⁰.

🖼 AI IMAGES OF THE DAY

All from the Lexica new model... insane (again)

👋 SEE YA