Lets make it

Turn viral videos into dinner. The AI Sous-Chef that extracts recipes from TikToks, scans your fridge, and fills your Instacart in one tap. 🍳

Comment

Inspiration

"We are living in the Golden Age of food content, but the Dark Age of actual cooking."

Every day, I watch dozens of cooking videos on TikTok and Instagram. I save them, thinking, "I'm definitely making those Birria Tacos."

But I never do.

Why? Because the friction is too high. A 60-second viral video has no ingredient list, no measurements, and no instructions. To cook it, I have to re-watch the video ten times, transcribe the ingredients, convert "a pinch" to grams, and then manually build a grocery list.

I realized that Social Media is an Inspiration Engine, not an Action Engine.

I built Let's Make to bridge that gap. I wanted a button that could turn a chaotic video into a structured, shoppable dinner in under 30 seconds.

What it does

Let's Make is an AI-powered "Sous Chef" that lives inside your iOS Share Sheet.

  1. The "Magic" Extraction: You see a video on TikTok. You tap "Share to Let's Make." Our AI analyzes the video audio and frames to extract a structured recipe (Ingredients, Steps, Times).
  2. The Fridge Scanner: You snap a photo of your open fridge. We use Computer Vision to identify what you already have and automatically cross it off your shopping list.
  3. One-Tap Shop: We deeply integrate with Instacart. One tap fills your cart with the missing ingredients.
  4. Hands-Free Cook Mode: A voice-activated interface designed for messy hands, so you don't smear flour on your phone screen.

How we built it

We built Let's Make as a "Headless" Intelligence Layer using Expo SDK 54 and React Native.

The Tech Stack:

  • Client: React Native + Expo Share Extension (Native Modules).
  • AI Pipeline: Supadata (Video Transcript) + Google Gemini 1.5 Pro (Structuring) + Gemini Vision (Fridge Scan).
  • Monetization: RevenueCat (Hybrid Consumable + Subscription Economy).

The Logic: Instead of complex math, we used a multi-modal approach. We treat the video as a data source. The AI listens to the audio (e.g., "add a cup of flour") while simultaneously looking at the video frames to verify the ingredient (detecting a bag of flour). This cross-referencing ensures we don't hallucinate ingredients that aren't there.

Challenges we ran into

  1. The iOS Memory Limit: iOS Share Extensions are ruthlessly killed if they exceed 120MB of RAM. Loading a full React Native engine + AI libraries caused instant crashes. We had to architect a "Headless Queuing System" where the Share Extension simply captures the URL and passes it to a background task (Expo TaskManager) to handle the heavy AI lifting.
  2. Hallucinations: Early versions of the AI would invent ingredients. If a video said "add the sauce," the AI would sometimes guess "Soy Sauce." We solved this by enforcing a Strict JSON Schema output from Gemini and cross-referencing against a "Master Pantry Database" of verified items.
  3. RevenueCat & Consumables: Managing a "Credit System" (1 Credit = 3 Recipes) required strict server-side validation. We used RevenueCat Webhooks to sync the ledger, ensuring that a user who bought a pack of credits on an iPad saw them instantly on their iPhone.

Accomplishments that we're proud of

  • The "Share Sheet" UX: It feels like magic. You never leave TikTok. You just tap a button, and a push notification tells you your dinner is ready to shop.
  • Visualizing the Invisible: Using Generative AI to create "Step-by-Step" images for videos that were too fast or blurry. We turn a low-quality 720p video into a 4K cooking guide.
  • The Hybrid Economy: Successfully implementing a business model where users can be "Casual" (buying $0.99 packs) or "Pro" ($9.99/mo) using a single RevenueCat implementation.

What we learned

  • Friction is the Enemy: We initially forced users to open the app to paste the link. Usage was low. Moving the experience to the Share Extension increased engagement by 400%.
  • Context is King: AI needs context. Simply transcribing the audio isn't enough; you need the visual frames to understand that "a pinch of heat" actually meant "Cayenne Pepper."

What's next for Let's Make

We see this as the "Commerce Layer" for the Creator Economy.

  1. Creator Partnerships: Launching "Chef Packs" where creators like Eitan can sell bundles of exclusive recipes as In-App Purchases.
  2. Smart Kitchen Integration: Sending cooking times and temperatures directly to smart ovens (GE/Samsung).
  3. Dietary AI: Automatically rewriting recipes. (e.g., "Make this Vegan" -> AI swaps Chicken for Jackfruit and adjusts cook times automatically).

Built With

Share this project:

Updates