How lakshmisravyavedantham/gurukul-ai works
Gurukul AI is a local-first educational video generator. Its main workflow takes a topic through research and script generation, creates scene images and narration, animates the scenes with a selected model, then assembles and subtitles the finished video. A separate agentic pipeline can generate animation variants, score and refine them, escalate models, and maintain a leaderboard. The README documents the model choices and optional cloud TTS/upscaling; the supplied excerpts do not show every internal call, so uncertain wiring is omitted.
Main entry points: app.py Gradio generator · agentic_pipeline.py Agentic pipeline · web_research.py Topic research · wan_animate.py Wan and LTX engines
Components 21
Topic and Script
- Gradio generator
app.py - Topic research
web_research.py - Scene script
generate_topic.py
Scene Media
- Scene images
app.py - Narration audio
app.py - Kokoro TTS
app.py - ElevenLabs TTS
app.py - Scene animation
app.py - Wan and LTX engines
wan_animate.py
Video Output
- Video assembly
assemble_video.py - Word subtitles
subtitles.py - Educational video
output/animated.mp4
Self Improvement
- Agentic pipeline
agentic_pipeline.py - Animation models
agentic_pipeline.py - Video critic
agentic_pipeline.py - Topaz upscaler
agentic_pipeline.py - Leaderboard dataset
agentic_pipeline.py
Other components
- Video creator
- DuckDuckGo and Wikipedia
- ComfyUI
- Topic input
Connections 21
- Video creator → Gradio generator: uses
- Topic input → Topic research: provides topic
- Gradio generator → Topic research: requests research
- Topic research → DuckDuckGo and Wikipedia: searches
- Topic research → Scene script: supplies facts
- Gradio generator → Scene script: generates scenes
- Gradio generator → Scene images: generates images
- Gradio generator → Narration audio: generates narration
- Narration audio → Kokoro TTS: uses local TTS
- Narration audio → ElevenLabs TTS: uses optional TTS
- Gradio generator → Scene animation: animates scenes
- Scene animation → Wan and LTX engines: selects engine
- Wan and LTX engines → ComfyUI: submits workflows
- Gradio generator → Video assembly: assembles clips
- Gradio generator → Word subtitles: adds subtitles
- Video assembly → Educational video: writes video
- Word subtitles → Educational video: writes subtitles
- Agentic pipeline → Animation models: generates variants
- Agentic pipeline → Video critic: scores outputs
- Agentic pipeline → Topaz upscaler: upscales best
- Agentic pipeline → Leaderboard dataset: records scores