Organise

Organise is where the app does its actual reasoning, and where you check it. Every decision on this screen is a proposal until you approve it.

To find out which files in a large library need attention before loading any of them here, run the library health check first. It’s a quick look from the files’ labels; Organise is the full check.

What it inspects, and why

Track labels are not evidence. A subtitle track named “English” is as likely to be a forced track covering one alien conversation as it is to be the full dialogue, and a media server that trusts the label will pick the wrong one every time.

So FileVerdict opens each track and looks at what is in it. For subtitles the strongest signal is caption density — how much text the track carries relative to the runtime. A full subtitle track has a cue every few seconds for two hours. A forced track has forty cues in total. That difference is unambiguous in a way the label never is.

Audio tracks are read from their channel layout, codec, language and title.

The track rows

Column	What it is
Track	The track’s own name, or (no name)
Type	Audio or Subtitle
Lang	The language recorded in the container
Verdict	What the track actually appears to be
Conf.	How strongly the evidence supports that
Action	Keep or Remove

Subtitle verdicts are Full, SDH, Foreign Language, On-Screen Text, On-Screen SDH Text and Commentary. Audio verdicts are Main, Alternate, Commentary and Descriptive. Subtitles explains what each one means and when it matters.

Suggested

A track whose confidence falls below Minimum confidence to auto-apply (Settings → Naming & Tracks) is badged Suggested rather than given a verdict to act on. A Suggested track is left exactly as it is, by Apply and by an encode, until you choose Keep or Remove.

The footer keeps a running count — “3 tracks waiting on a decision” — and the Add to queue button repeats it, so you cannot walk past unresolved tracks by accident.

Raising the threshold means more tracks wait for you. Lowering it means fewer do, and more decisions get made on thinner evidence.

With more than one file loaded, the filter button above the list, Show only files that need a review, hides every file with nothing left for you to decide. What stays is the files with a Suggested track, plus any still being read. The number on the button is how many that is, and the button turns purple while it’s on. Click it again to see everything.

It only changes what the list shows. Every file is read and judged the same way whether it’s on or not, and you can turn it on before or after adding files. A file drops out of the list once nothing in it is left to decide, including right after you decide its last Suggested track.

A track is also marked Suggested, however confident its verdict, when its label disagrees with what its contents read as: titled “SDH” but reading as plain dialogue, or flagged forced but full of it. Either the file is mislabelled or the reading is wrong, and a glance at the track settles which.

Reviewing with the keyboard

For a long list of Suggested tracks, you can work through them without the mouse:

Key	Does
↑ / ↓	Move to the previous or next track
N	Jump to the next track waiting on a decision
K	Keep the selected track
R	Remove the selected track

Keeping or removing a Suggested track moves straight on to the next one waiting, so a season’s worth is a run of key presses. On a track you’ve already decided, K and R change it and stay put. The keys do nothing while you’re typing in a field, while a window is open, or on a file that’s in the queue. Every one of them is also a click: the keys are only a quicker way.

From label

With Scan and read subtitle track contents switched off, a subtitle track’s verdict comes from its label, its flags and the file header rather than its contents, and shows from label in place of a confidence figure. See Scanning speed for how those verdicts are made.

Not scanned

A track can be listed without a verdict, showing not scanned. Two reasons:

• It is already set to be removed, so reading it would have cost time to inform a decision that was already made.
• Scan and read subtitle track contents is switched off and the track’s label does not say what it is — see Scanning speed.

Either way the track is not counted against you and nothing about it is rewritten. Each file offers a read N unscanned link to read them on demand; it works per file rather than per track because the expensive part is opening the file at all, so once that is paid the remaining tracks are nearly free.

The evidence panel

Click a track and the right-hand panel shows its name, its verdict, its Keep/Remove action, and Evidence — the actual statements behind the call. Cue counts, caption density, whether OCR ran and on how many samples, flags found in the container.

This is the point of the whole screen. A verdict you disagree with can be checked rather than argued with: if the evidence is wrong, the verdict is wrong, and you can see exactly which step went astray. Override the verdict or the action from the same panel — your choice is what gets written.

An audio track also shows Encoded as: what Compress will make of it, such as “Opus · 5.1 · 256 kbps”, and whether that came from one of your preferences or a one-off override. It is shown here, not set here — the setting itself lives in Compress’s Audio tab, explained in Audio encoding — so that a track’s language, its verdict and what it will become are all readable in one place. A track set to remove says so instead, since nothing is written for it.

Looking at the subtitles themselves

Inspect Subtitles opens a sampled view of the track’s actual cues with their timings, and lets you reclassify it from there. For an image-based track these are the OCR results, which is also a quick way to see whether OCR is working at all.

Edit subtitles opens the full cue list for a text-based track (SRT, VTT, ASS) and lets you delete lines — the usual target being a “Subtitles by …” credit or an advert baked into a downloaded file. Image-based tracks cannot be edited this way; there is no text in them, only a picture per cue, until they’re converted to text.

Converting picture subtitles to text

Convert to text, in a picture subtitle track’s panel (PGS or VobSub), reads every line into a text track that takes the picture track’s place when you Apply or encode. Picture subtitles make Plex transcode on many devices; text ones don’t. It keeps the track’s verdict, name, flags and timing, takes about half a minute for a film, and reads English only for now. Text recognition can misread a word now and then, so the new track is worth a look with Edit subtitles. See Subtitles.

Syncing subtitles that are early or late

Sync subtitles fixes a track whose lines show up before or after they are spoken. It plays a minute of the video with the track on top, and you move the track until the lines match the speech. The shift is written when you apply or encode, and the track’s row shows it until then. See Syncing subtitles.

Copying your changes to a whole season

The episodes of a season usually have the same tracks, so a change you make to one is usually right for all of them. Instead of making it on every episode, make it once and copy it.

This saves you clicks, not scanning time. Every episode is still read in full, and FileVerdict judges each one’s tracks for itself. Copying only repeats the changes you made. To make scanning itself quicker, see Scanning speed.

How it works

1. Change something on one episode: set its SDH track to Remove, say, or rename its English subtitles.
2. A Copy your changes box appears at the top of the track panel: “The tracks you changed are also in 12 other episodes, which don’t have your changes yet.”
3. Click Copy your changes to all 12. One click makes every change you made to that episode, on every track at once, whichever track you had selected.

The box only appears once you’ve changed something, and only while there’s an episode your changes would change. Once they all have them, it just says “Copied to 12 episodes.” Change something else and it offers to copy again.

You don’t have to wait for the whole season to be read. Only episodes that have finished reading count, and the box says how many are still being read. Copy to the ones that are ready, and the box offers the rest as they finish.

What gets copied

Only what you changed, each change to the same track in the other episodes:

• Keep or Remove, including a Suggested track you settled. A track still Suggested on this episode is left as it is on the others.
• track names, languages and flags
• a verdict you changed

Everything else in the other episodes stays as it is: their other tracks, changes you made to them yourself on other tracks, their filenames, library folder ticks, subtitle timing and any subtitle files you added. If an episode has a change of its own to the same thing, FileVerdict names it and asks before replacing it.

An episode you copied to doesn’t offer those changes on to others. Change something on it yourself, and it will.

What counts as the same track

A change goes to the track in each other episode with the same type, language and format, that FileVerdict read the same way: the English subtitles it read as SDH, say, or the English commentary audio. Names don’t matter, and neither does the order tracks are listed in.

It’s matched track by track, so what else an episode has doesn’t matter. Releases often carry a different mix of other-language subtitles from one episode to the next. Those don’t stop a change to the English tracks reaching every episode.

The reading is part of it so that a change never lands on a track that’s really something else. Take The Glass Harbour, where a few episodes have a forced subtitle track for dialogue in another language:

• A change to the forced track goes to the other episodes’ forced tracks. Episodes without one are unaffected by it.
• A change to the full English subtitles goes to every episode’s full English subtitles, forced track or not.
• An episode where the forced track reads as full subtitles has two tracks read that way, where the others have one. A change to the full subtitles can’t tell which of its two is the one, so it changes neither rather than guess. That episode’s forced track is usually Suggested too, so Show only files that need a review will show it.

Renaming

The Rename suggestion row shows the filename the file would be given. Click it to see the candidate matches from your metadata source, pick a different one, or type the name yourself.

Renaming is independent of track edits — you can rename without touching a single track. See Renaming files.

With Organise into library folders on in Settings, the Destination bar shows the file’s place in your library, and the pane has a Move into the library folder box. It starts ticked. Untick it to leave just this file out of the library, this time: the setting stays on for every other file, and the choice isn’t remembered. See Library folders.

Adding subtitles

Under each file’s tracks, + Add subtitle lets you pick a subtitle file from disk: .srt, .ass, .ssa, .vtt or .sub. The language is guessed from the filename.

The subtitle is classified on arrival and appears as a normal track row with the same verdict, evidence and Keep/Remove controls as any embedded track. It is muxed into the file when you apply.

Applying

Three buttons, and they do genuinely different things:

Apply without encoding writes the track edits and the rename immediately. No re-encode, no queue — seconds, not hours. For a pure clean-up this is the one you want.

Add to queue sends the files to the Queue to be re-encoded with your Compress settings, with the track edits applied as part of the same pass. With a single file and nothing else queued, the button reads Start encode instead: it does the same thing, and the encode starts at once.

Encode without organising re-encodes using the original tracks and filename, discarding every decision on this screen, and leaves the file out of the library. It exists for the case where you opened Organise, changed your mind, and just want the file smaller.

Danger

Track removal is not reversible — no copy of the removed track is kept. With Keep the original file on (the default), edits are written to a new file and your source is untouched, which is the safety net to leave on until you trust the results.

With Keep the original file off, the file is read again once it has been written, so its rows show what is in it now rather than what was there before.