Dictionary
The Dictionary lists every word spoken in your project's analyzed videos, with each place it's said. Search for a word, click it, and you hear it right there. Paste a script and it tells you which words the project can supply and which it can't.
It's built from the word-timed transcripts that Redub produces on import. Nothing is stored separately. Bring in another video and its words show up, delete a sequence and they go.
Opening it
The Dictionary belongs to the whole project, not to one sequence, so there are two ways in:
- On the Sequences page, click Dictionary.
- With a sequence open, pick the Dictionary tab at the end of the tab strip.
Opened from the Sequences page, a Sequences button in its header takes you back.
Only analyzed videos contribute
A word has to come from a sequence made with New Redub whose analysis has finished. Edit timelines, storyboards, and an Assistant cut that only gathers other sequences have no transcript of their own, so they add nothing. When two sequences share the same source video (a duplicate, say), that video is counted once.
Finding a word
Type in Search words. Words that start with what you typed come first, then words that merely contain it. Accents are optional, so cafe finds "café".
Two buttons set the order: A-Z sorts alphabetically, Count puts the most used words first. The number at the end of each row is how many times the word is said across the project.
Click a word to open it. You get one row per place it's said, showing the time, the sequence, and the media file name. Opening a word also plays its first take, so you hear it straight away. Click any other row to play that one instead. A word said hundreds of times shows its first 200 places with a row underneath to show more.
The preview player
The player plays just the word you picked, from its source media, then stops. Under it:
- Play replays or pauses. With the pointer over the player, Space does the same.
- Loop repeats the word until you turn it off.
- Play with surrounding context adds a second and a half either side, so you can hear the word inside its sentence and judge how cleanly it would lift out.
- Open in sequence leaves the Dictionary and opens the sequence in Redub, with the player and timeline parked on that moment. Use it when you want to edit, not just listen.
An audio-only source plays the same way, with the file name shown in place of a picture.
Checking a script
Type or paste a script into the Script panel and click Process, or press Ctrl+Enter. The script comes back marked up:
| Marking | Meaning |
|---|---|
| Teal word | Said somewhere in the project. |
| Teal word with a brass underline | Part of one continuous take: these words are said back to back, in this order, in a single video. |
| Red word with a wavy underline | Not said anywhere in the project. |
The line above the script sums it up: how many words are available, how many are missing, and how many takes cover how many words.
Click a word to hear it. A word inside a take plays the take from that word to its end, which is usually what you want, since one continuous read always cuts better than the same words gathered from five places. Any other word plays if it's only said once, and asks you which place you mean if it's said several times. The small number beside a word always opens the full list of places, even for a word inside a take, in case you want it from somewhere else.
Edit script takes you back to the text. The check reruns by itself whenever the Dictionary changes, so a video that finishes importing can turn red words teal while you watch.
How takes are found
For each script word, Kitlig tries every place it's said and keeps the longest stretch where the video goes on to say the same words your script does, with no pause longer than about 0.8 seconds. It works left to right, so the take under a word is always the longest one that starts at that word. It doesn't search for the smallest possible number of takes overall.
Matching ignores capitals and punctuation, and treats a hyphenated compound like "twenty-five" as two words. It matches exact words only: "run" won't find "running", and "25" won't find "twenty-five".
Inside the Assistant Editor
In the Assistant Editor, the toolbar's Dictionary button docks a compact Dictionary beside the cut preview, with Words and Script tabs. The preview keeps playing when you open or close it.
Here a click works differently. Instead of playing the source video, it moves the cut preview to the point where that moment plays in your cut. Places that are in the current cut carry an in cut badge. Click one that isn't, because its take is unused or that stretch was trimmed away, and a bar says so and offers Show in Dictionary, which opens the full Dictionary on that word.
The dock starts with only the videos this cut draws from. Tick Whole project to search everything.
The dock only looks things up. It never changes your cut.
Exporting
Two buttons in the Dictionary header save it to a file.
Export JSON writes a single dictionary.json: every word, the file it's in and when it's said, plus the sequences behind those words. It carries no media. It's readable by other tools, and it's also importable, as long as the other project can get at the videos.
Export Archive… writes one .zip holding that same file together with the media. Before you choose where to save, the dialog shows how many sequences and files are going in and how large the result will be. Tick Source videos and analysis only (smaller) to leave out dubbed audio, stems, and extracted clips. Sequences imported from a smaller archive keep every word and can still be re-analyzed, but they open without their dubs.
The archive export runs as a job, so you can keep working while it writes. Cancelling it removes the partial file.
Archives are as large as the videos in them
Media is stored as is, not recompressed, since video and audio are already compressed and squeezing them again costs time for almost nothing. An archive over 4 GB is written in the ZIP64 format. Current unzip tools read it, but very old ones may refuse it.
Importing
Import… takes a dictionary .zip or a dictionary.json. Kitlig reads the file first and shows what would happen before anything is written: each sequence, its word count, and one of these:
| Shown | Meaning |
|---|---|
| media in archive | The video comes out of the zip. |
| media already here | This project already holds the identical file, so it's reused and not copied again. |
| no media | The words come in, but there's nothing to play yet. |
| already here | The project already has this sequence, so it's left alone. |
Click the Import button (it shows how many sequences it will add) and the new sequences are created as real Redub sequences. They appear on the Sequences page, open in Redub, and can feed an Assistant cut, exactly as if you'd analyzed the videos here. Import into an empty project and you get the same Dictionary you exported. Import into a project with its own videos and the new words are added to what's there.
Importing the same file twice adds nothing the second time. A sequence also counts as already here when the project has one with the same name over the identical video.
When it finishes, a notice sums up what was added. View report lists what was copied, what was reused, and anything that was skipped or failed.
What doesn't travel
Assistant draft history isn't carried over, so an imported Assistant cut arrives as a plain Redub sequence ready for a fresh brief. Storyboard cast and location links to Kitlig Assets are dropped, and so are image-sequence and segmentation assets.
Importing without the media
A dictionary.json on its own has no videos in it. If the project already has the identical files, they're matched and linked without you doing anything. For the rest, the import dialog offers Look in a folder…. Point it at a folder holding the videos and Kitlig matches them by file name, then checks each one against the original before using it, so a different file that happens to share a name is never linked by mistake.
Anything still not found comes in as words only. It's fully searchable, and the script check counts it, but the player has nothing to play. Pick one of its words and the player shows Locate media…. Choose the file and the sequence is linked to it from then on.
Link the same file the words were timed against
Word timings are measured against one exact file. If you locate a different file, Kitlig warns you first. A re-encode of the same video will usually still line up. A different edit of it won't, and words will play at the wrong moments.
Older sequences
Sequences analyzed before word-level timing existed have no per-word timestamps, so they add nothing to the Dictionary. When the project has any, a button appears in the Dictionary header saying how many older sequences need updating. It opens the same Update Sequences dialog described on the Video mode overview, which rebuilds the timings from the analysis already saved, without re-running any detection.