Viral Hooks
Trending short videosExplaining an AI food model
S

@softgirlnocode

Instagram · Talking Head Edit

@softgirlnocode I’ll share the website with you just comment “butter.” Regardless of how you feel about the practicality of this whole model, I actually think this is a really cool free recipe generator. You know when you have random leftovers in the fridge and ingredients that don’t really make sense together? You can throw them into the website, and it helps generate recipe ideas around them. The interesting part is how the model understands which ingredients “go together.” It maps each ingredient into a 300-dimensional vector based on two things: 1. Co-occurrence in recipes, basically what ingredients humans already tend to use together. 2. Shared flavor compounds from FlavorDB and FlavorGraph so it’s not just guessing based on recipes, it also considers food chemistry. #foodie #sidequest #tech #llm

View original on Instagram

The edit1 shot · 7 text
Shots Text1:16

Performance

Views
372.8K
Likes
15.4K
Comments
9K
Engagement
6.6%
Outlier
9.9×
Length · 29 May 2026
1:16

Video #37450 · Talking Head Edit

Explaining an AI food modelfrom @softgirlnocode

[01]

Overview

The creator explains a new AI model called Epicure, which understands flavor by analyzing 4 million recipes and food chemistry. She demonstrates how it can generate unique ingredient pairings and recipes, citing examples like white chocolate with caviar and passion fruit with garlic. The creator concludes by questioning the impact of this technology on chefs and noting Silicon Valley's growing obsession with 'taste'.

[02]

Simple outline

  1. 1Introduce novel technology
  2. 2Demonstrate with surprising example
  3. 3Validate with extreme user test
  4. 4Pose provocative question
  5. 5Deliver insightful conclusion

[03]

Related videos

239.6K23.7×AI recipe study reveals world map@harnidhk · Breakdown2M1.3×Explaining new AI marketing tools@rpn · Breakdown390K10.8×Smart kitchen gadget product demo@fdotinc · Product Demo724.2K5×Robot vs human chef taste test@planetmoney · Explainer1M1.6×Explaining an AI manipulation experiment@wetclaude · Case Study Breakdown85.4KAI video editing process breakdown@solocontent.studio · Tutorial / How To154.7K4.9×Explains 3D printing problem, solution@fdotinc · Opportunity Explainer1.4M134.1×AI tool demo with examples@kallaway · Product Demo623.7KSpeaker explains AI's social media future@rpn · Breakdown5.3M1506.9×AI tool enhances chaotic food fight@adobefirefly · Product Demo805.8K34.3×Illustrated breakdown of tech controversy@nataliefratto · Case Study Breakdown144.2K3.8×Tutorial for AI app self-improvement@angus.sewell · Tutorial / How To

[04]

Classification

Format
Talking Head Edit
Concepts
Opportunity ExplainerBreakdown
Elements
B-Roll CutawayImage OverlayText Overlay
Account
Personal Brand
Topics
TechFoodScienceCooking

[05]

Hook analysis

What stops the scroll in the first 3 seconds

Title hook · 15%

AI can understand

Verbal hook · 70%

This is AI understanding flavor.

Visual hook · 15%

A split-screen composition featuring a complex data visualization graph on top and a low-angle shot of the creator outdoors reaching for a keyboard on the grass.

[06]

Hook template

Steal this for your own video

There's a new [technology or tool] that actually understands [human sense or experience] — and it's going to change [industry or profession] forever.

  • — There's a new AI model that actually understands flavor — and it's going to change how chefs cook forever.
  • — There's a new algorithm that actually understands emotion in music — and it's going to change how songwriters create forever.
  • — There's a new scanning tool that actually understands pain patterns in the body — and it's going to change how physical therapists treat patients forever.

[07]

Hook and payoff

The promise the opener makes, and how the video pays it back

Hook

A bold, high-concept statement that immediately defines the topic and sparks curiosity with an intriguing visual.

Payoff

A final, insightful thesis that reframes the technology's purpose, offering a novel perspective that leaves the viewer feeling informed and intellectually satisfied.

[08]

Narrative framework

The structural pattern this video runs on

Framework

The Counter-Intuitive Discovery

Introduce a new technology or concept, demonstrate its power through a series of increasingly surprising and counter-intuitive examples, and conclude with a broader reflection on its implications.

  1. 1

    The Bold Proclamation

    State the video's core thesis in a single, declarative sentence to establish the subject matter instantly.

  2. 2

    The Technical Foundation

    Briefly explain the mechanism, data, or scale behind the thesis to build credibility and establish its power.

  3. 3

    The Practical Pivot

    Directly address the audience's likely question ('So what?') and transition from the abstract concept to its tangible, real-world application.

  4. 4

    The Shocking Example

    Present a counter-intuitive, almost unbelievable example of the concept in action to break viewer expectations and prove its non-obvious power.

  5. 5

    The Social Proof Escalation

    Introduce a third-party discovery or user-generated example that is even more extreme, validating the concept and escalating the intrigue.

  6. 6

    The Live Demonstration

    Act as the audience's proxy by testing the most extreme example, showing the process and revealing a surprisingly plausible result to resolve curiosity.

  7. 7

    The Provocative Question

    Shift from demonstration to implication by posing a challenging question to a specific group, raising the stakes of the discussion.

  8. 8

    The Contextual Zoom-Out

    Place the specific topic within a larger, relevant industry trend to give it greater significance and timeliness.

Themes

  • AI Innovation
  • Future of Food
  • Data Science
  • Creativity vs Technology

Techniques

  • B-Roll Cutaway
  • Direct-to-camera address
  • Use of social media screenshots

Post caption

I’ll share the website with you just comment “butter.” Regardless of how you feel about the practicality of this whole model, I actually think this is a really cool free recipe generator. You know when you have random leftovers in the fridge and ingredients that don’t really make sense together? You can throw them into the website, and it helps generate recipe ideas around them. The interesting part is how the model understands which ingredients “go together.” It maps each ingredient into a 300-dimensional vector based on two things: 1. Co-occurrence in recipes, basically what ingredients humans already tend to use together. 2. Shared flavor compounds from FlavorDB and FlavorGraph so it’s not just guessing based on recipes, it also considers food chemistry. #foodie #sidequest #tech #llm

Text overlays

  • title · beginning

    AI can understand

  • title · beginning

    Flavour

  • informational · beginning

    Launching our new paper on arXiv: we trained the largest multilingual food model ever built. 4.1M recipes. 7 languages. 1,790 ingredients. 300 dimensions. All of human cooking compressed into 2 megabytes.

  • contextual · middle

    white chocolate caviar

  • informational · middle

    Standouts: soy sauce + vanilla (111 shared compounds!), passion fruit + garlic, coconut + beef.

  • informational · middle

    Pan-Seared Sea Bass with Passionfruit and Garlic Butter

  • informational · end

    WHY TECH BROS ARE NOW OBSESSED WITH TASTE

Transcript

This is AI understanding flavor. Someone just trained the largest LOM on food by compressing over 4,000,000 recipe from around the world into this vector space, which is this dense map of ingredient relationship. And you're probably wondering what the heck is that no for? Well, you can generate recipe and ingredient pairing that's backed by culture science that you probably never know existed before. But did you know why chocolate and caviar is a great pair? Apparently, they share flavor compounds. I don't know who's eating that, but apparently, the molecules are vibing. You can test this on their website, and this comment make me laugh. Someone said they've been on the site and found 25 ingredient pairs that are chemically extremely close, but absent from every recipe on earth, like soy sauce and vanilla and passion fruit and garlic. No, of course, I had to test to see what recipe can you come up with that. And this is what I good, 10 seared sea bass with passion fruit and garlic butter. Honestly, I would try that. If you're a chef right now, are you furious that AI is claiming that it knows taste and is coming for your creative monopoly? Because Silicon Valley has been obsessed with taste for the past eighteen months, and it's just funny there's an AI model now that's literally built for taste. I do think there's something real there because different cuisine will rotate and pair ingredients in completely different way. And a model that's trained on all of that can suggest directions that you've never been culturally exposed to.