Skip to content
 
 

Repository files navigation

AgenticAI Project | PixFrame AI

An end-to-end AI-powered animated video generation system that converts a single prompt into a complete short film using modular LLM agents.


image image Screenshot 2026-05-03 205943

Overview

This project implements a multi-phase agentic pipeline:

  1. Story & Script Generation — Prompt → structured story, scenes, characters
  2. Audio Generation — Dialogue + character voices + background music
  3. Video Composition — Scene visuals + animation + A/V sync → final video
  4. Web Interface — UI to run pipeline, track progress, and preview output
  5. Edit Agent — Natural language edits with versioning and undo support

Tech Stack

Layer Tools
Backend FastAPI
Frontend Next.js / React
Agents LangGraph / LLM APIs
Audio ElevenLabs / Coqui / Bark (TTS)
Video FFmpeg / MoviePy
Image Generation Stable Diffusion / APIs
State Management JSON + version snapshots

Quick Start

Backend

pip install -r requirements.txt
python -m backend.app

Frontend

cd frontend
npm install
npm run dev

Features

  • Prompt → full animated video pipeline
  • Scene-wise generation (modular phases)
  • Real-time progress tracking (Phase 4)
  • Phase-level re-run support
  • Edit agent with:
    • Natural language edits
    • Targeted regeneration (audio / video / script)
    • Version history (v1, v2, v3…)
    • Undo / revert system

Project Structure

agents/          # Phase-wise agents (story, audio, video, edit)
backend/         # API + orchestration layer
frontend/        # Web UI
mcp/             # Tooling layer (LLM, audio, video, etc.)
state_manager/   # Versioning + snapshots
data/            # Outputs and assets
shared/          # Schemas and utilities

Notes

  • All phases communicate via a shared JSON state schema
  • Each phase is modular and independently testable
  • Every run/edit creates a new version snapshot

About

PixFrameAI | Agentic AI Project

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages