Skip to main content

Match Clips To Reference Framing

This workflow solves a finishing problem that is tedious to do by hand: you have a reference version of a cut — an approved render, an offline edit, a client-supplied master — and your own timeline with the original clips, and the framing does not match. Position and scale were adjusted somewhere along the way, and now someone has to eyeball every shot.

A typical case is a vertical teaser that was edited, approved, and rendered in 9:16 while the original footage is 16:9. When you rebuild the edit with the original media, every shot needs the same crop choice: the right part of the wider source frame must be shifted and scaled into the vertical frame. This workflow tries to recover those baked framing decisions and turn them back into editable Motion Position and Scale values.

The library workflow Match Clips To Reference Framing does this automatically: it extracts frames from both versions with FFmpeg, visually matches each eligible target clip against the reference, and writes the recovered Position and Scale values back onto your clips.

This workflow does not identify missing source clips or reconstruct edit timing. Start with Reconstruct An Edit From A Reference when the original shots are not yet on the timeline, then use this page for the separate framing pass.

This is also a good example of what Runtime Packs make possible — the analysis runs on real external tools (FFmpeg and Python with OpenCV), well beyond what timeline scripting alone can do.

New to Automation Agent?

This recipe uses Automation Agent for Adobe Premiere. If you are here for the workflow solution first, that is fine. Use the product overview and Set Up Your AI for the setup path.

When This Is Useful​

  • an approved reference render exists, and the conformed timeline must match its framing
  • a vertical, square, or otherwise reframed version was approved, and you need to rebuild it from source footage with a different aspect ratio
  • reframing decisions from an offline edit need to be recovered onto the original media
  • a client or colleague supplied a flattened version, and you need the punch-ins and repositions as editable Motion values
  • you want framing as explicit Position/Scale values you can inspect and adjust, instead of baked-in pixels

What You Need First​

  1. A sequence with clips to adjust on one video track. V1 is the default target.
  2. One or more online reference media files or renders on a different video track. V2 is the default reference. The reference must show the approved framing and use the same pixel dimensions as the active sequence. If the reference raster differs, pre-render it at the sequence dimensions first; the workflow stops before applying changes rather than guessing how Premiere conformed the image. For each eligible target clip, the workflow uses the first overlapping reference clip and analyzes only the overlap. Do not use a nested sequence, offline clip, or generated-media item as the reference: the workflow reads the underlying reference file with FFmpeg, so use a rendered video file instead.
  3. The FFmpeg and Python Runtime Packs. The panel's preflight reports what is missing and points you to the Runtime Packs tab.

No Workflow AI provider is needed — the matching is deterministic image analysis, not model judgment.

How To Run It​

  1. Open the Library tab and load Match Clips To Reference Framing from Examples > Finishing.
  2. Select the target clips you want to process. Selected target clips are the recommended default; choose All clips on target track in the dialog when you intentionally want a full-track pass.
  3. Run it. Choose the reference and target video-track numbers, the target clip scope, the temporary analysis folder, and whether to keep extracted frames and matcher reports. Track numbers use Premiere's labels, so V1 = 1 and V2 = 2.
  4. The workflow extracts overlap frames from the chosen target and reference tracks with FFmpeg, matches them visually, and applies one static Position/Scale update per matched target clip.

The values are applied through Premiere's language-independent built-in Motion match name, so the workflow also works on localized Premiere installations.

An optional adaptive keyframe mode samples near the clip boundaries and across the overlap, rejects isolated inconsistent matches, and writes linear Position/Scale keyframes when it finds a consistent change. It checks the midpoint between keyframes and only adds another keyframe when that measurably improves the reconstructed framing. A final simplification pass removes redundant keys and keeps Position or Scale static when only the other parameter changes. Matcher samples are recorded in sequence time and converted to each clip's source-time coordinates before Premiere Motion values are read or written, so clips with non-zero timeline starts or source in-points keep their keyframes inside the visible clip range.

Expected Result In Premiere​

Each eligible target clip that overlaps a reference gets updated Motion Position and Scale values that reproduce the reference framing. Target clips without any reference overlap are skipped and reported. A selected-only run replaces this workflow's markers only inside the processed clip ranges; markers for unselected target clips remain. An all-clips run replaces all previous markers from this workflow. Temporary analysis files are removed after a successful run by default.

The workflow also creates review markers with match status and confidence information. Those markers make it clear which clips matched cleanly and which clips need manual attention because the target source and reference do not share enough visible image content.

During longer runs, the Run bar reports preparation, analysis chunks, application, and finalization. The completed run leaves a nonmodal result summary beside the Run control and writes the same structured result to the console.

The result stays fully editable: the recovered framing is normal Motion values on your original clips, not a re-render.

Limitations And Review Points​

  • The workflow recovers Position and Scale only. Direct reference media must match the active sequence raster exactly. Differently sized or differently shaped references must be pre-rendered at the sequence dimensions because Premiere can conform them outside the readable Motion Scale. The workflow can account for Position and Scale already set on a compatible reference clip, and for shifts or scale changes that are baked into the rendered reference pixels.
  • Rotation is not supported. Baking Rotation into the reference does not make it recoverable; it usually makes the visual match worse because the matcher is not solving for rotated images.
  • Reference Motion Anchor Point should usually stay neutral. If an anchor-point change is already baked into pixels as a simple image shift, the workflow can only treat that as Position; it will not reconstruct Anchor Point as a separate Motion value.
  • Masks, adjustment layers, grading changes, and other effects are not reconstructed. They may still be present in the reference pixels, but the output remains Position/Scale values on the target clips.
  • The default run applies one static update per clip; framing that animates within a clip needs the adaptive keyframe mode. Adaptive refinement is bounded to avoid runaway analysis time or excessive keyframe counts, so very fast or highly irregular motion can still need manual adjustment.
  • Review the result shot by shot before conforming further. Visual matching is robust, but extreme scaling, heavy grading differences, rotation, very flat content, or unrelated target/reference imagery can reduce match confidence — the markers and matcher reports help when something looks off.