reels-af

reels-af is an open-source AI multi-agent system designed to automate the creation of viral short-form videos for platforms like Instagram Reels, TikTok, and YouTube Shorts. It orchestrates multiple AI agents that work together to research trending content, write scripts, generate voiceovers, source or create visuals, and assemble finished videos — all with minimal human intervention.

The project is developed by Agent-Field and is hosted on GitHub at github.com/Agent-Field/reels-af. Built in Python, it leverages a modular pipeline architecture where each step of video production is handled by a dedicated agent, making it highly customizable and extensible.

• Multi-Agent Pipeline - Separate AI agents handle research, scripting, voiceover, visual selection, and video assembly.
• Trend Research - Automatically discovers trending topics and viral content patterns for better engagement.
• AI Voiceovers - Generates natural-sounding voice narration using text-to-speech models.
• Visual Sourcing - Pulls relevant images, clips, and B-roll automatically based on script content.
• Auto-Captioning - Adds synchronized captions and subtitles optimized for short-form platforms.
• Batch Production - Can produce multiple reels in a single run with different topics or angles.
• Docker Support - Ready-to-use Docker configuration for easy deployment and scaling.

The reels-af pipeline processes content through several stages:

Stage	Agent	Output
Research	Trend Agent	Topic brief with viral hooks
Script	Writer Agent	Optimized short-form script
Voiceover	Audio Agent	AI-generated voice narration
Visuals	Media Agent	Matching images and video clips
Assembly	Editor Agent	Final rendered video with captions

The system uses the AgentField control plane to coordinate agent communication and task distribution.

Clone and install:

git clone https://github.com/Agent-Field/reels-af.git
cd reels-af
pip install -r requirements.txt

Copy the example environment file and configure your API keys:

cp .env.example .env
# Edit .env with your API keys for LLM, TTS, and media services

Docker

docker build -t reels-af .
docker run -it --env-file .env reels-af

Basic command:

python -m reel_af --topic "Your topic here"

From an article:

python -m reel_af --article "https://example.com/article"
# or
python -m reel_af --article-file ./content/article.txt

The pipeline accepts either a topic (generates both script and visuals from scratch) or an article URL/path (repurposes existing content into a reel format).

reels-af is free and open-source software. The ~$0.10 per reel cost comes from third-party API usage fees for LLM inference, text-to-speech generation, and media sourcing. Users bring their own API keys, so actual costs depend on the chosen providers and the complexity of each video.