what you wish giphy was
- Executive Summary Cooked. is a high-speed iMessage App Extension that uses AI to generate hyper-personalized memes. By combining the recipient's contact photo with the immediate chat context, it creates one-of-a-kind "burns" that feel organic to the conversation.
Name: cooked.
Aesthetic: TBD, modern playful young
Tagline: "Cooked in a click."
App Icon Concept: A minimalist, glowing orange frying pan icon on a matte black background.
A. The "Delayed Wit" Problem User Need: Users want to be the "funny one" in the group chat, but by the time they find a good GIF on Giphy, the conversation has moved on.
Solution: Cooked. generates a relevant meme in <2 seconds using the last thing copied to the clipboard.
B. The "Generic Content" Problem User Need: Standard memes feel stale. Users want content that feels "inside joke" level.
Solution: By face-swapping the recipient into the meme, the content becomes 10x more engaging than a generic reaction GIF.
C. The "Privacy Friction" Problem User Need: Users are wary of AI apps "stealing their face data" or reading their private messages.
Solution: Cooked. uses a "Privacy-by-Action" model. It doesn't read the chat; it only processes what the user explicitly Copies. It uses contact photos already stored on-device to build trust.
Module 1: Contact & Face Intelligence Entry Point: MSMessagesAppViewController detects the active thread.
Logic: Request CNContactStore access. Pull thumbnailImageData for the recipient.
AI Prep: If no contact photo exists, the app defaults to a high-quality "Face-less" caricature or prompts for a quick snap.
Module 2: Contextual Analysis (LLM) Input: UIPasteboard.general.string (The copied "dumb" text).
LLM Role: Claude/GPT-4o identifies the type of roast (e.g., "The Over-Explainer," "The Late Arriver," "The Bad Take").
Output: Generates 3 custom captions and selects the most "iconic" base meme template for the face-swap.
Module 3: Generation & Insertion Generation: A server-side API (FastAPI/Node) handles the face-swap using InsightFace.
Optimization: Return a small, high-compression .gif or .heic to ensure instant loading in iMessage.
UX: Single-tap insertion into the message bubble.
Trigger: Friend sends a message. User long-presses and selects Copy.
Open: User taps the Cooked. icon in the iMessage tray.
Preview: The app displays three "Roast Previews" (Face-swapped GIF + AI Caption).
Send: User taps the best one. It’s inserted into the input field.
Result: User hits the send arrow. Cooked.
TTR (Time to Roast): Seconds from app open to meme sent (Target: < 3s).
Viral Coefficient: Number of new installs attributed to "Sent via Cooked." watermarks.
Retention: Percentage of users who "Cook" at least 3 friends per week.
Next Step for Claude Code "Now that the PRD is set, let's start with Module 1. Claude, can you write the Swift code for an iMessage Extension that requests Contact permissions and retrieves the profile picture of the person currently being messaged in the active conversation?"