😺Deepfakes just got 10x better

PLUS: 7 o3-mini prompts to learn anything...

Welcome, humans.

Alphabet shares fell 9% after announcing a revenue miss and a cool $75B in new AI capex yesterday. Google’s CFO said they ended the year with “more demand than we had available capacity”, which could explain both those issues.

Google also used the earnings call to flex their Gemini 2.0 Flash and Flash Thinking models are “some of the most efficient out there”, even compared to DeepSeek and R1. And yet, they STILL can’t meet demand. Classic Jevons paradox, am I right?

AMD, NVIDIA’s rival, also reported earnings, and even though its data center sales doubled in 2024, it still missed expectations.Data centers just can’t catch a break these days. One minute there’s not enough, the next minute, still not enough—I mean, what are we going to do, give them more than two minutes to catch up?‍

Here’s what you need to know about AI today:

BytDance demo’d the ultimate deepfake tool.
Cloudflare adds Adobe image verification across its network.
Google dropped AI weapons restriction from its principles.
Robot startup Figure AI leaves OpenAI, claims its own breakthrough.

Advertise in The Neuron here.

Deepfakes are about to get out of control…like 10x better out of control.

‍Ever wished you could make that old photo of your grandma sing “Happy Birthday”? Or turn your LinkedIn headshot into a TED talk? Or, IDK, create a realistic version of yourself to attend Zoom meetings while you’re sipping Mai Tai’s in Honolulu?

‍Well, ByteDance (TikTok's parent company)’s new AI model can do exactly that, and TBH, it’s sorta terrifying.

‍OmniHuman-1 takes any single photo—selfies, portraits, full-body shots, even cartoons—and makes them move realistically based on audio or video inputs. And unlike previous models, it works with pretty much any kind of image (paper here).

Here are 10 incredible examples—can you guess which is our favorite? Shockingly, it’s NOT this clip of Taylor Swift singing a Naruto theme song, or THIS clip of Jensen Huang rapping…but that was a close second.

What makes OmniHuman-1 special is its versatility:

Creates eerily realistic gestures and expressions.
Handles both singing and talking naturally.
Even works with non-human objects.

Think of it like a universal translator between still images and motion. You feed it a photo and some audio, and it figures out how to make that photo move naturally with the sound.

‍See, China isn’t just leading the open-source AI race for reasoning models… with Tuesday’s release of OmniHuman-1, and previous leading video models like Hunyuan, Kling, and MiniMax, it’s clear the country is heavily focused on video content, too

‍And this is just one piece of the synthetic media explosion:NVIDIA-backed Synthesia just raised a new round of $180M that values the company at $2.1B and now has 60K customers. It and HeyGen (who raised $60M) are the two leading avatar startups, and for good reason—they look good.Gone are the days of wonky static avatars; these tools now provide motion control and other features that make their avatars more realistic than ever.Synthesia gets its realistic voices from ElevenLabs, the leading voice AI who just raised its own round of $180M at a valuation of $3.3B. There’s also Cartesia, who makes an incredibly life-like voice AI called Sonic (demo here).

‍All of these companies want to become more lifelike. You look at their roadmaps, and that’s basically what’s coming—moving AI from the uncanny valley to the make money valley.The problem? The most profitable use for life-like avatars is deepfakes, either used to influence elections or scam people and corporations out of billions. And if scammers get access to tools like OmniHuman-1 to play with? I think Darth Vader said it best…

FROM OUR PARTNERS

This is the Biggest Breakthrough in Next-Gen Tech Since iPhone—but their round is closing in less than a week.

Elf Labs won historic rights to iconic characters like Cinderella, Snow White, Sleeping Beauty, and The Little Mermaid. Now they’re transforming the $2 trillion media industry by leveraging cutting-edge patented technology to bring them to life like never before.

Think VR—without headsets—and AI powered talking toys. More exciting: they’re on the brink of a game-changing product launch—one that has the potential to reshape their future. They’ve recently opened an investment round to the public and investors who act now will lock in their stake before share prices jump.

Fundraise closing early—Deadline: February 12.
Industry-shifting product/partnership announcement imminent.
$2.00 per share—last chance before next valuation update.
Already 1,200 investors—and growing!

But time is running out. Invest Before the Round Closes 2/12Disclosure: This is a paid advertisement for Elf Labs’ Regulation CF offering. Please read the offering circular at elflabs.com--*Please note that the investment opportunity featured in our newsletter is presented solely as a sponsorship and does not constitute an offer, solicitation, or recommendation to buy or sell any securities.

Prompt Tip of the Day

‍Robin Delta just dropped 7 killer prompts for o3-mini + search to master any topic without spending a dime on courses.

‍Our favorite? The “Socratic Enlightenment” prompt transforms o3 into your personal Socrates. Instead of just info-dumping, it guides you through deep, probing questions that actually stick in your brain. Super useful for challenging concepts.

‍Pro tip: After (Bill and Ted voice) So-crates-GPT guides your learning, run the “Teach-back Mastery” prompt to teach the concept you just learned back to o3. o3 will spot any gaps in your understanding, and help you patch them up. Party on, dudes!!!

Treats To Try.

Tool Finder lets you search and filter through thousands of software tools to find exactly what you need for your work.
Chatbase enables customer service AI to take actions like processing refunds and scheduling meetings, not just chat.
Icon.me turns your video library into complete social media ads by automatically matching clips to your scripts and handling the rest.
Sendbird answers your customers' questions instantly across all channels.
Zinc lets you search and order from major retailers through one simple API, so your apps can handle e-commerce automatically (starts at $50 prepaid: 1¢ per product search, $1 per order).
Swatle analyzes your project tasks and automatically breaks them down into manageable subtasks with time estimates.
Synthesia 2.0 is Synthesia’s enterprise platform that lets you build personal avatars as well as editable screen recordings and turning docs into videos.

See our top 51 AI Tools for Business here!

Around the Horn.

This is Dylan Patel of Semi Analysis, a very credible source—though keep in mind, it’s still just a rumor. Full source here.

CloudFlare announced it will integrate Adobe’s content authenticity framework into its image service as a one-click solution to verify the authenticity of images as they move through the web—which is a big deal, since CloudFlare manages ~20% of all internet traffic.
Google updated its AI principles and removed a section titled “applications we will not pursue” that included a pledge not to use AI for weapons.
Figure (AI robot company) left its collaboration agreement with OpenAI because it made a “major breakthrough on fully end-to-end robot AI, built entirely in house”—it also plans to ship 100K robots over the next four years.

FROM OUR PARTNERS

This FREE guide will help you win with AI…

Top sales teams are saving 10+ hours per week with AI—learn their secrets in our new, TOTALLY FREE guide (created in partnership with Attention), complete with 10+ ready-to-use sales prompts that anyone can use.Get the free guide.

Where do you #Neuron?!

Submit where you #Neuron here for a chance to be featured in our newsletter next week! Rules to get featured:

Monitors / phones in a unique location.
If you include your face, we’ll protect your identity.
Cats = heavily encouraged.
Dogs = case by case basis.

Jas from Solihull, UK: I start the day by reading the Neuron.. the only newsletter I read in full!

Saurabh from Pune, Maharashtra: When it’s time to TASTE what NEURON is cooking, you know it’s gonna be a recipe for brainy chaos—served with a side of ‘Did I just forget why I walked into this room?’