Organize Video References and Study Them Frame by Frame

By refern | Last updated: July 2026

The short answer: refern indexes video files in the same local library as your images, so a clip is searchable, taggable, and openable right next to the stills you saved from it. You can step a video frame by frame, mark an A and B point to loop a section, drop labelled markers on the moments you care about, and save any frame as a PNG reference.

If you study motion, this removes the split most reference workflows still live with: images in a library, footage in a folder somewhere else, and a separate video player open whenever you actually want to look at it.

Before you start

Video indexes the same way images do, so there is no separate import mode to learn. You need:

• refern installed with a workspace open (Windows, macOS, or Linux)
• A folder containing video files, either already inside your workspace or ready to be added as a content root
• Clips in one of the indexed formats: mp4, mkv, mov, avi, webm, flv, wmv, m4v, 3gp, or m2ts

One note on scope: refern indexes images, video, and its own canvas files. It does not index audio, fonts, PDFs, 3D models, or layered source files like PSD. If your reference folder holds a mix, the video and images come across and the rest is skipped.

Step 1: Get your footage into the library

Point refern at the folder your clips already live in. It indexes them in place and never copies or moves the originals, exactly as it does with images.

Each video gets a poster-frame thumbnail, so a folder of footage reads as a grid of frames rather than a list of filenames. Hovering a clip in the grid scrubs through it, which makes finding the right take a matter of sweeping across a row instead of opening files one at a time.

If a clip is in a container your system cannot play directly, refern converts it to a playable copy automatically. You do not have to transcode anything by hand first.

Step 2: Find the clip you want

Video carries its own search operators on top of everything you already use for images:

• duration: filters by length, so duration:>=1m finds your long takes
• fps: filters by frame rate
• codec: filters by codec
• sort:duration orders results by length, longest first, or add -asc for shortest first

These compose with the operators you already know, so type:video tag:walk-cycle duration:<=30s is a valid single query. If you want a refresher on the syntax, the guide to refern's search operators covers the full set.

Any of these searches can be saved as a smart folder that keeps itself current. "Every reference clip under ten seconds" becomes a folder you never have to maintain.

Videos also work in groups, carry ratings and tags like anything else, and get suggestions from the on-device Auto Tagger, which samples frames from a clip rather than skipping it.

Step 3: Study the clip frame by frame

Open a video and it fills the same viewer your images use, with the same shortcuts and the same metadata sidebar. The player adds what motion study actually needs:

• Frame-by-frame stepping in both directions
• A storyboard preview on the timeline, so hovering the scrub bar shows you where you are heading
• An audio waveform and a filmstrip, for finding a beat by eye or by ear
• Playback speed, loop, and volume, plus resume where you left off
• Timecode display modes and a keyboard shortcut sheet
• Zoom and pan, and picture-in-picture when you want the clip floating over your drawing app

This is the part that usually forces a second application. A dedicated player gives you frame stepping but knows nothing about your library; your library gives you organization but plays video like a web page. Doing both in one window means the clip you are studying is already tagged, already grouped with the stills you pulled from it, and already one keystroke from the rest of your reference.

Step 4: Mark the moments you care about

Drop a labelled marker at any timestamp. For a walk cycle that might be contact, down, passing, and up. For an acting reference it might be the beat where the expression turns.

Markers are more than bookmarks:

• Jump between them instead of scrubbing blindly
• Save any marker's frame as an image, which lands in your library as an ordinary reference with your tags on it
• Export your markers and the current section to .otio and .edl timeline files, which your editing application can read

That last one matters if the reference work feeds a real edit. The marks you made while studying travel to the timeline instead of being retyped.

Step 5: Pull stills and palettes out of the footage

Two things you can do from any frame:

Save or copy the frame. The current frame saves as a PNG into your library, tagged and searchable like any other still. This is the honest way to build a practice set from footage: pull the frames you want, then work from those.

Pull a color palette from the frame, then search your library by those colors. If a shot has exactly the light you are chasing, you can go from that frame to every other reference you own in the same palette without exporting anything first. The color search guide covers how that scoring works.

Step 6: Loop or export a section

Mark an A point and a B point to define a section. From there you can:

• Loop it, which is the core study move for a short beat
• Export it as an .mp4 clip
• Cut it losslessly, a stream copy with no re-encode that keeps the source container and quality
• Save it as an animated .webp loop, capped at 30 seconds
• Trim the original, which is destructive but undoable

The lossless option is the one to reach for when you are pulling a reference beat out of a large file and do not want a generation of quality loss for the privilege.

Step 7: Put the clip on a canvas

Drag a video onto a refern canvas and it plays in place, with scrubbing and frame stepping available from the selection panel. The clip sits on the same board as your stills and notes rather than in a window behind it.

You can also build a board straight from a video, or from its markers, which turns a marked-up clip into a shot breakdown in one step.

How this compares

Two honest notes on that table. Eagle indexes video and a great deal more besides, including fonts, audio, and 3D; its format breadth is the widest of any tool in this category and refern does not try to match it. And PureRef's GIF playback is genuinely good for what it is, with frame scrubbing and frame extraction since version 2.0. The gap is that a GIF is not a reference clip, and PureRef has no library to put one in.

Common pitfalls

A clip has no thumbnail yet. Poster frames are generated by the same background pipeline that thumbnails images. On a large batch of new footage they fill in progressively while you keep working.

A video will not play. If the container is one your system cannot decode, refern converts it to a playable copy automatically. If a file is in a format outside the indexed list, it is not in the library at all, which is a different problem with a different fix: convert it to one of the supported containers.

You want a practice set from footage. Pull the frames you want as PNGs first, then build the session from those images. That is the reliable path, and it leaves you with reusable stills.

Your loop export is capped. Animated .webp loops stop at 30 seconds. For anything longer, export the section as a clip instead.