Text-to-Music 🎶 and text-to-4D 🩻

Hey folks! This is Ben's Bites - we're the lead singer and you're the backing vocals. We both need each other to make this thing sound beautiful.

I'm flying to Austin tomorrow, get there in time for dinner and I've got to amuse myself Sat eve and all day Sunday. If you have any recommendations - tweet them at me @bentossell

Let's get to it!

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 20k subscribers? Check out our sponsor options.

Join Discord

🤌 Ben's Picks

1/

MusicLM: Generating Music From Text, via Google.

Holy shit. It's text to music! Go and listen to the examples on the link - its so impressive.

I really struggle to find music I like (old-school hip-hop) because it's just not maaayyaayyyayyyde that wayyyyayyyayyy any more. (sorry about that). Democratising the creation of music is huge. It'll be interesting to see what's built on top of this in the future. Low-fi playlists, white noise, bathroom music, anything!

This article provides a list of musical prompts and the corresponding audio elements that can be used to create a variety of musical pieces.

Different musical prompts can be used to create different audio elements, audio elements can be used to create a variety of musical pieces, and different musical pieces can be created with different audio elements.

2/

MAV3D: Text-To-4D Dynamic Scene Generation, via Meta.

This article presents MAV3D, a method for generating three-dimensional dynamic scenes from text descriptions.

MAV3D uses a 4D dynamic Neural Radiance Field (NeRF) to generate dynamic video output, it does not require any 3D or 4D data and the T2V model is trained only on Text-Image pairs and unlabeled videos, and it is the first to generate 3D dynamic scenes given a text description.

3/

When in Rome, do as the Romans do... The modern-day version is “when in meetings, do as the Japanese do”. Inemuri is the Japanese art of “sleeping whilst being present”, and now you can practise it too with Supernormal’s AI tool for summarising meetings, which just received $10m in funding and has over 50,000 users at some big firms. It’s setting itself apart from the competition with the ability to extract actions and critical decisions.

4/

Is this the first case of generative AI being used in Hollywood? The Oscar-tipped film Everything Everywhere All At Once reportedly used Runway’s generative video editing tool in their VFX team. If you’ve seen this insane film with insane special effects and insane cinematography, you could totally see how this could be the case.

There are a lot of text2video-type tools and tech coming out at the moment, but Atin Gupta outlines how they all fall quite short of the real deal: a mish-mash of videos-with-no-sound, human-portrait-videos-with-sound, and video editors, but nothing yet combining all 3. Bring on the first person who can nail that combo!

🛠️ Cool Tools

• Watch Now - Let AI find your next movie or show. (link)

• Outset AI - Ask AI how it can help your business. (link)

• Writing Mate - AI communications helper and writing companion powered by GPT-3, ChatGPT. (link)

• Bible GPT - Describe your situation to get advice from the Bible. (link)

• Voice Pen AI - Convert audio content into blog posts. (link)

• Promptify - Use GPT or other prompt based models to get structured output. (link)

• Slite introduces private beta for QnA based AI search across their workspaces. (link)

• SmoothTalker - A quick AI pickup line generator. (link)

• Konjer - A library full of books you can talk to. (link)

• WebChatGPT v.2 - Augment ChatGPT prompts with relevant results from the web. What’s new - A prompt editor. (link)

🤓 Miscellaneous

• How we’re approaching AI-generated writing on Medium. (link)

• The world’s first robot lawyer isn’t a lawyer, and I'm not sure it’s even a robot. (link)

• ChatGPT can’t be credited as an author, says the world's largest academic publisher. (link)

• Is ChatGPT a step toward human-level AI? Podcast with Meta chief AI scientist Yann LeCun. (link)

• A new form of UI is emerging with prompt-driven design. (link)

• The story so far - AI makes for strange bedfellows. (link)

• AI code assistants head to head. A comparison of top AI coding tools. (link)

• Hume AI aims to endow AI with EQ and concern for human well-being. (link)

• Scale - the $290M/year Mechanical Turk of machine learning. (link)

• First look - ChatGPT + WolframAlpha. (link)

• Actionable AI. How AI will become more actionable. (link)

🎓 Learn

• How to create a realistic video character using only free AI tools. (link)

• How to turn any video clip into an AI masterpiece with today's Runway Academy. (link)

🔬 Research

• Improving statistical fidelity for neural image compression. (link)

• On the importance of noise scheduling for diffusion models. (link)

• MusicLM from Google Research - Generating music from text. (link)

• Unsupervised 3D animation of non-rigid deformable objects. (link)

• Cut-and-LEaRn (CutLER) - A simple approach for training unsupervised object detection and segmentation models. (link)

• DetectGPT - Detecting samples from pre-trained LLMs using the local curvature of the model's log probability function. (link)

• Understanding finetuning for factual knowledge extraction from language models. (link)

• MAV3D (Make-A-Video 3D) - Generating three-dimensional dynamic scenes from text descriptions. (link)

• Simple Diffusion: End-to-end diffusion for high resolution images. (link)

📰 Unclassifieds

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 20k subscribers? Check out our sponsor options.

🖼 AI images of the day

🤗 Share Ben's Bites

Send this to 1 AI-curious friend and receive my AI project tracker database!

or copy/paste this link: https://bensbites.beehiiv.com/subscribe?ref=PLACEHOLDER

⭐️ How did we do?

⭐️ REAL REVIEWS