Guide
Complete documentation for the VIDIM AI Broadcast Analysis System. Learn how to load MXF files, run the pipeline, browse clips, adjust IN/OUT points, and export results.
Workflow Overview
VIDIM follows a 6-step workflow. Steps 1-2 are automated, steps 3-4 are interactive review, and steps 5-6 produce output.
1.Select MXF files — Use the file browser to select one or more broadcast recordings.
2.Run the pipeline — Click Start. All 11 stages run sequentially (5-12 min per heat).
3.Browse clips — Review auto-generated clips in the clip browser grid.
4.Adjust IN/OUT — Fine-tune clip boundaries using the frame-accurate viewer.
5.Export — Export approved clips with metadata to the output directory.
6.View report — Review the structured JSON report with statistics and timings.
VIDIM WORKFLOW
1
LOAD MXFSelect broadcast MXF files from local storage
2
RUN PIPELINE11-stage analysis: frames, OCR, state machine, clip creation
3
BROWSE CLIPSReview auto-generated clips in the clip browser grid
4
ADJUST IN/OUTFine-tune clip boundaries using frame-accurate viewer
5
EXPORTExport approved clips with metadata to output directory
6
VIEW REPORTReview structured JSON report with statistics and timings
Steps 1-2 are automated. Steps 3-4 are interactive review. Steps 5-6 are output.
Total pipeline time for a typical heat: 5-6 minutes
Documentation
01→02→03→04→05→
Pipeline Reference
11 sequential stages, 3 AI models, processing performance, and discipline codes. Includes a live UI preview.
Clips & Frame Viewer
7 clip types with scoring thresholds, the clip browser grid, frame-accurate viewer, and keyboard shortcuts.
Reports & Export
Report contents breakdown, export formats (JSON, CSV, TXT, PDF), and the reports UI with expandable detail views.
AI Agents & IRC
9 agents with roles and descriptions, real-time status grid, and the YUHEX IRC network with 6 channels.
API & FAQ
16 REST endpoints, WebSocket events, and answers to common questions about formats, performance, and capabilities.