# social_recipes

> Self-hosted tool that extracts recipes from TikTok, YouTube, and Instagram videos using AI, then imports them into Tandoor or Mealie.

- **Category:** Recipe Management
- **Licence:** MIT
- **Latest version:** v1.9.10
- **Open source:** oss
- **Languages:** Typescript, JavaScript, Python, HTML, CSS
- **Platforms:** Docker, Docker Compose, Build from Source
- **Tags:** Automation, Video Processing, Food, Pwa, Instagram, Youtube, LLM, Gemini, Openai, Cooking, Docker, AI, Python, Flask, Transcription, Nutrition, Self Hosted, Whisper, Tiktok, Mealie, Recipe Extraction, Tandoor

## Description

Pick-a-Recipe is a self-hosted Python web app that pulls videos from social media sites via yt-dlp, turns the audio into text with Whisper, reads on-screen details using vision models, builds structured recipes through configurable AI backends (OpenAI, Gemini, Anthropic, or any OpenAI-compatible service), and sends the results to a self-hosted Tandoor or Mealie server. It ships a web interface with live progress, user accounts (local or OIDC), PWA installation on phones, and optional nutrition data from Open Food Facts. You can run it in Docker (amd64/arm64) or install it manually with Python 3.11 and FFmpeg. All settings are adjusted in the web UI.

## Links

- [Homepage](https://github.com/pickeld/social_recipes)
- [Docker](pickeld/pick-a-recipe)
- [GitHub Repository](https://github.com/pickeld/pick-a-recipe)
- [Docker Hub](https://hub.docker.com/r/pickeld/pick-a-recipe)

---
Source: https://selfhost.directory/project/socialrecipes