Skip to main content

Build A Reusable Auto-Conform Source Index

Feature films, episodes, and campaign source reels can take much longer to scan than a short trailer reference. The library workflow Build Auto-Conform Source Index lets you perform that source scan once—often before the edit session or overnight—and reuse it for every later Reconstruct Edit From Reference run that uses the same media.

The index contains classical visual descriptors, not copied video frames. It is written in bounded chunks, so an interrupted scan keeps completed work and resumes from the missing chunks the next time you run it.

Run The Library Script​

  1. Install the FFmpeg and Python/OpenCV Runtime Packs.
  2. In the Project panel, select the direct source footage that later reference edits may use.
  3. Open Library > Examples > Finishing and run Build Auto-Conform Source Index.
  4. Choose a fast local Auto-Conform workspace.
  5. Keep Reuse valid indexes selected unless you deliberately want to rebuild the selected media.

No active sequence or reference video is required.

After a successful run, the workflow leaves a nonmodal completion summary beside the Run control and writes the same structured result to the console. It includes the number of processed sources, the reusable workspace, and the durable report paths. The workflow writes both a detailed JSON report and a human-readable .txt processing-time report. A repeated run can finish almost immediately when every selected source already has a valid cached index.

Use The Same Workspace Later​

Choose the same workspace in Reconstruct An Edit From A Reference. The two workflows share this layout:

Auto-Conform Workspace/
  cache/      reusable source indexes
  reports/    durable source-index and conform reports
  runs/       temporary uniquely named run folders

Keep the workspace on fast local storage even when the original media lives on a server. The descriptors are read repeatedly during matching, while the source video is accessed mainly during initial indexing and final frame refinement.

Reuse, Resume, And Rebuild​

  • Reuse valid indexes (Recommended) uses complete cached indexes, resumes partial ones, and rebuilds only missing or invalid chunks.
  • Rebuild selected sources removes only the cache identities belonging to the currently selected source items and scans those items again. It does not clear unrelated campaign indexes.
  • Replaced or modified source files receive a different PATH_SIZE_MTIME identity and therefore do not silently reuse stale descriptors.
  • The cache is bounded and applies least-recently-used pruning while protecting every source selected by the current run.

Each run writes a JSON report containing the selected sources, cache hits, rebuilt source count, chunk/sample counts, timing, cache paths, and memory usage. The matching .txt file summarizes wall-clock processing time, cache reuse, and the number of indexes built or rebuilt for production benchmarking.

Current Limitations​

  • Sources must be direct, online media files readable by FFmpeg.
  • Sequences, generated media, offline clips, and ambiguous multicam interpretations are not indexed.
  • The workspace is local to the computer and media paths used to create it. Portable shared indexes across workstations are not promised yet.
  • Cancellation takes effect between bounded workflow operations; it cannot interrupt every individual FFmpeg decode instantaneously.