Social Media to Mealie

Have you found a recipe on social media and don’t want to write it out yourself? This tool lets you import recipes from videos directly into Mealie.

Tested social media platforms:

• Instagram
• TikTok
• Facebook
• YouTube Shorts
• Pinterest

Other sites may work as well, since the tool uses yt-dlp to download videos. If you encounter issues with other websites, please open an issue.

Note: If you receive a BAD_RECIPE error, it may be due to Mealie’s recipe parsing. If you find a better prompt or solution, feel free to open an issue or PR!

Features

• Import posts into Mealie with a link and a click
• iOS Shortcut v0.3 for easy importing
• PWA support, to allow sharing links to the app from mobile devices for quick importing.

Screenshot

Requirements

• Mealie 1.9.0+ with AI provider configured (docs)
• Docker

Deployment

Docker Compose

1. Create a docker-compose.yml file based on the docker-compose.example.yml in the repo and fill in the required environment variables. If you prefer having them in a separate file you can create a .env file based on the example.env.

2. Start the service with Docker Compose:

   docker-compose up -d

Docker Run

docker run --restart unless-stopped --name social-to-mealie \
  -e OPENAI_URL=https://api.openai.com/v1 \
  -e OPENAI_API_KEY=sk-... \
  -e TRANSCRIPTION_MODEL=whisper-1 \
  -e MEALIE_URL=https://mealie.example.com \
  -e MEALIE_API_KEY=ey... \
  -e MEALIE_GROUP_NAME=home \
  -p 4000:3000 \
  --security-opt no-new-privileges:true \
  ghcr.io/gerardpollorebozado/social-to-mealie:latest

Local Development

1. Clone the repository and install dependencies:

   git clone https://github.com/GerardPolloRebozado/social-to-mealie.git
   cd social-to-mealie
   npm install

2. Create a .env file based on example.env and fill in your values:

   cp example.env .env

3. Create an empty cookies.txt file (required by the volume mount if defined in docker-compose.yml):

   touch cookies.txt

4. Build and start the container locally:

   docker compose build
   docker compose up -d

   The app will be available at http://localhost:4000.

5. To view logs:

   docker logs -f social-to-mealie

Tip

In order to be able to install the PWA the app needs to have HTTPS, you can use a reverse proxy like caddy or nginx, but be careful as opening this app to the internet as it will allow anyone to submit recipes to your Mealie instance. I recommend adding a small log in or IP whitelist

Environment Variables

Variable	Required	Default	Description
OPENAI_URL	Yes	https://api.openai.com/v1	URL for the OpenAI API or a compatible one
OPENAI_API_KEY	Yes	—	API key for OpenAI or a compatible one
TRANSCRIPTION_MODEL	No	whisper-1	Whisper model to use, required when the local one is not filled
LOCAL_TRANSCRIPTION_MODEL	No	—	Model ID from hugging face to use for local audio to text transcription, required when the provider doesn't support transcriptions API
TEXT_MODEL	Yes	gpt-5-mini	Text model to use for recipe generation
MEALIE_URL	Yes	http://localhost:9000	URL of your Mealie instance
EXTERNAL_MEALIE_URL	No	-	Optional, this will be the URL used when showing the result of the recipe, not needed if the MEALIE_URL is already a public accessible URL
MEALIE_API_KEY	Yes	—	API key for Mealie
MEALIE_GROUP_NAME	No	home	Mealie group name
EXTRA_PROMPT	No	—	Additional instructions for AI, such as language translation
YTDLP_VERSION	No	latest	Version of yt-dlp to use
PORT	No	4000	Host port to expose the app on
COOKIES	No	—	Cookies string for yt-dlp to access protected content NAME=VALUE

Tested AI providers compatibility

• OpenAI
• GroqAI

Partial support

Because theese providers don't support the transcriptions API it requires LOCAL_TRANSCRIPTION_MODEL to be set, recommended model: Xenova/whisper-base, you can use any model that is compatible with the ONNX runtime from hugging face

• llmstudio
• ollama

Some recommended models for local AI are:

• qwen3-vl:8b
• gemma-3-12b

If you want to use another model the model needs to have tools support and vision capabilities. If you want better results use the same models but in larger variants.

It can work with any other provider that is compatible with the OpenAI API, if you find any issues please open an issue.