CaseStudies

MassGen Video Recording and Editing

Status: 📋 Planned
Version: Future
Last Updated: November 15, 2025

Overview

Auto-generate case study videos by running commands with recording, editing (speed up, add captions, highlight logs), cutting unnecessary parts, and producing 1-minute demo videos automatically for showcasing MassGen capabilities.

Description

Goal

Automate the entire video production pipeline for MassGen case studies: from running the command to publishing a polished 1-minute demo video, eliminating manual video editing work.

Key Features

Automated Recording
- Record entire MassGen execution session
- Capture terminal output, logs, and visual output
- Track important events (agent responses, tool calls, results)
- Support both terminal and browser recording
Intelligent Editing
- Speed up boring sections (compilation, waiting, repetitive output)
- Cut unnecessary parts (errors that were recovered, redundant logs)
- Highlight key moments (final answers, important decisions, insights)
- Add captions for important commands and outputs
- Picture-in-picture for multi-agent coordination
Content Analysis
- Video understanding to identify key frames
- Log analysis to find important events
- Automatic chapter markers
- Generate video description and keywords
Production Quality
- Add intro/outro sequences
- Background music (optional)
- Smooth transitions between sections
- Professional color grading and effects
- Export in optimal format and resolution
Multi-Format Output
- Full recording (for documentation)
- 1-minute highlight reel (for social media)
- 30-second teaser (for Twitter)
- GIF animations (for docs/GitHub)
- Tutorial segments (for YouTube)

Workflow

Run MassGen Command
    ↓
Record Session (asciinema/OBS)
    ↓
Analyze Recording (identify key moments)
    ↓
Edit Video (speed up, cut, add effects)
    ↓
Generate Multiple Formats
    ↓
Publish (YouTube, Twitter, Docs)

Testing Guidelines

Test Scenarios

Short Task Recording (5 min execution)
- Task: Simple research query
- Test: Record, edit, produce 1-min video
- Expected: Captures key moments, smooth pacing
- Validation: Video is watchable and informative
Long Task Recording (30 min execution)
- Task: Complex multi-agent workflow
- Test: Handle long recording, intelligent time-lapse
- Expected: Condense to 2-3 minutes without losing narrative
- Validation: All major steps visible, progression clear
Multi-Agent Recording
- Task: Parallel agent execution with coordination
- Test: Show multiple agents working simultaneously
- Expected: Picture-in-picture or split-screen layout
- Validation: Easy to follow, coordination visible
Error Recovery Recording
- Task: Task with failure and recovery
- Test: Show error briefly, then skip to recovery
- Expected: Error visible but not dwelled on
- Validation: Maintains flow, shows resilience
Caption Accuracy Test
- Task: Recording with important commands/outputs
- Test: Auto-generate captions for key moments
- Expected: Captions are accurate, well-timed, readable
- Validation: Human review of caption quality
Full Pipeline Test
- Task: Run case study, produce video automatically
- Test: End-to-end automation with no manual intervention
- Expected: Publication-ready video in <10 minutes
- Validation: Video quality suitable for public sharing

Quality Metrics

Technical Quality:

Resolution: 1080p minimum
Frame rate: 30fps minimum
Audio quality: Clear, no artifacts
Compression: Balanced quality/size

Content Quality:

Narrative clarity: Easy to follow
Pacing: Not too fast or slow
Information density: Key points visible
Engagement: Holds attention

Production Value:

Transitions: Smooth and professional
Captions: Readable and well-placed
Effects: Subtle and helpful
Branding: Consistent with MassGen identity

Validation Criteria

✅ Full automation: command → published video
✅ 10:1 compression ratio (10 min → 1 min) without losing key info
✅ Human evaluation: Video quality >7/10
✅ Caption accuracy >95%
✅ Processing time <10 minutes per video
✅ Videos suitable for public sharing (YouTube, Twitter)

Implementation Notes

Technical Requirements

Recording:

asciinema for terminal (from v0.1.14)
OBS Studio for screen capture
Browser automation recording
Event logging for synchronization

Video Understanding:

Multimodal models (v0.1.3) for frame analysis
Log parsing for event detection
Importance scoring for key moments

Editing:

FFmpeg for video manipulation
Python libraries: moviepy, opencv
Automated editing scripts
Caption generation

Planning Mode:

Break editing into phases
Coordinate multiple tools
Handle >10min processing time
Progress tracking and reporting

Configuration Example

video_production:
  recording:
    mode: auto
    capture: screen_and_terminal
    fps: 30
    resolution: 1920x1080
  
  editing:
    speed_up_threshold: 5  # Speed up if no activity for 5s
    max_duration: 60  # Target 1 minute
    cut_errors: true
    add_captions: true
    highlight_key_moments: true
  
  output_formats:
    - full_recording  # Complete session
    - highlight_reel  # 1 minute
    - teaser  # 30 seconds
    - gif_animations  # Key moments
  
  production:
    intro_outro: true
    background_music: subtle
    branding: massgen

Execution Command

# Record and auto-produce video
massgen --config case_study.yaml \
  --query "Research AI trends and write report" \
  --record \
  --auto-edit \
  --output-video ./demos/ai_trends.mp4

# Batch process multiple recordings
massgen-video-edit \
  --recordings ./recordings/*.cast \
  --template highlight_reel \
  --output ./videos/

Terminal Evaluation (v0.1.14 planned) - Session recording with asciinema
Multimodal Video Analysis (v0.1.3) - Video understanding
Automation Mode (v0.1.8) - Structured output for analysis

References

Key Value: Reduce video production time from hours to minutes, enabling rapid case study publication and increasing MassGen visibility through high-quality demo videos.

Target Output: Professional demo videos similar to existing MassGen case study videos on YouTube, but produced automatically.