Skip to content

Captions: generate, verify and edit

The three pipelines, the subtitle editor, tracks, gaps and highlights: everything between raw dialogue and approved text.

The three pipelines

A pipeline decides where the words come from. Choose it on the Generate page, in the Audio card, under Pipeline. Extraction, Translation and Scripts all lead to the same editor and the same Ignite.

PipelineInputOutputUse it when
ExtractionThe dialogue on the active timelineWord-timed subtitles in the spoken languageYou want captions of what was actually said.
TranslationThe dialogue, plus a source and target languageTranslated subtitles with their own timingYou need a second-language version of the same cut.
ScriptsThe dialogue and an approved scriptThe script, not the spoken wording, decides the textA client-approved script must appear verbatim.

Scripts needs Furnace. Scripts is a Pro and Max pipeline, and its option appears only while Furnace is on.

With Scripts selected, press + in the Audio card to add Script 1, Script 2 and so on, and paste one script into each. Every card must be filled or removed before Start. A clip slot is one source clip's span of the timeline: Scripts takes one script per slot, the slot decides the pairing, and the card order does not have to match the source order.

Starting a pass

  1. Choose the range

    Caption the whole timeline, or mark an In and an Out to work on one section. In a Resolve session, set them on the connected Resolve timeline with Resolve's own I and O. In an Ember project, press I and O on Ember's timeline, and press Alt+X to clear both. Ember only reads the range you give it.

  2. Set the language

    Pick the source and target languages for Translation; Extraction and Scripts detect the spoken language themselves. Source and target must differ.

  3. Set up the optional extras

    Turn on Furnace for a rough cut before captions, Shuffle for Wave and Formation styling, and Highlight for accented words.

  4. Press Start

    Start runs the pass and uses credits, so run it once with the settings you want. Changing a caption's look later does not require another Generate.

Ember's language picker covers 13 spoken languages, plus Auto. Languages with language-aware orthography keep accents, capitalization and punctuation intact through client review.

Reading the subtitle editor

The editor shows the cut as a timeline of subtitle blocks. Each block is one caption on screen; each block holds the words that were spoken while it is up.

Blocks

A block has a start, an end and the text it shows. Its edges are the visible geometry: once you move or trim a block, that is the timing the timeline uses.

Words

Every word keeps its own time inside the block. Word timing drives word-by-word reveals, Highlights and sound cues.

Tracks and lanes

Captions can share a timeline in several placement lanes. Ember starts with four subtitle lanes and adds more as your blocks need them.

Waveform

The audio waveform sits under the blocks. Use it to confirm that a line lands on the speech instead of trusting the transcript alone.

Editing text and timing

Inline edit
Double-click a block, or press Return with a block selected, to type in it. Ctrl+Enter (Cmd+Enter on macOS) or clicking away saves; Enter starts a new line; Esc discards.
Split at playhead
Press H, or choose Split at Playhead from the block menu. It cuts one block into two at the playhead, keeping the words on either side.
Add block
Press A, or choose Add Subtitle Block on empty timeline space. It creates an empty block at the playhead for text that was missed.
Delete blocks
Press G, or choose Delete Block from the menu. It removes the selected captions and their words.
Move words
Select words and press Left or Right, or use the word menu. It sends words to the previous or next block when a line break is in the wrong place.
Nudge
Alt+Up and Alt+Down move selected blocks one lane up or down. Ctrl+Left and Ctrl+Right move them in time.
Highlight word
Press X with the pointer over a word in the timeline or a Subtitles card. The word menu offers the same action. It marks a word for the accent colour without changing the Highlight percentage.
Detonate
Press R, or choose Detonate from the block menu. It breaks a block into word-level blocks when a fast exchange needs tighter reveals. This is not the Detonate Wave on Magic Shuffle.
Switch Style
Press S, or use the Style row at the top of the block menu. It opens the Style picker for the selected blocks.
Enable / disable
In an Ember project, press D or use the item's own switch. Enable / disable takes captions and other timeline items out of the picture while keeping them in the project.

All of these are undoable. If a pass or an edit does not feel right, undo in Ember rather than rebuilding the captions.

Checking timing and coverage

  • Play the cut with the waveform visible. A block that starts late or ends early is easier to hear than to read.
  • On Pro and Max, ask your assistant to list the pauses with no caption, or to read the transcript in order with each word's timeline position. On Lite, read the waveform and the gaps between blocks yourself: a pause with no block over it, or a block that starts after the speech, is visible there.
  • Generate holds a caption through a short pause between neighboring captions; longer silences are left open so captions do not linger.
  • Check words that matter: names, brands, product terms and numbers. Fix those by hand instead of paraphrasing the speaker.

Lines, tracks and the safe area

Ember builds readable lines instead of dumping a full transcript into one block. Four controls shape them: Format, Characters, Highlight and End Buffer.

Format
Sets Lines, Case and Punctuation. Lines allows one line or two.
Characters
The most characters a line may hold. Drag it to Word for one word per caption.
Highlight
The share of blocks that receive an accented word, from 0 to 100 percent.
End Buffer
Holds a block a little longer after its last word.
  • A Formation is one of Magic Shuffle's arrangements of neighboring blocks. It skips multiline blocks, so allowing two lines changes what Shuffle can do with those captions.
  • A Style has three roles: Regular for ordinary captions, Hook for the opening line Cast writes, and Special for blocks you assign a Special Style. The Styles page covers all three.
  • Captions stay inside a fixed safe margin shared by Ember's preview, its own render and the Resolve output.
  • Placement lanes let captions stack or sit above and below each other without colliding.
  • Switch Style gives selected captions another Style. Ignite Override in the Inspector lets one caption's Font rows differ from the rest of its lane.

Highlight

Highlight finds the words worth accenting, such as names, brands, technical phrases and anchor words, and colors them with the accent of the caption Style. Set a percentage from 0 to 100 to decide how many blocks receive one; 0 turns it off.

Highlight Punctuation, in Settings > Editing, is off by default. Turn it on when punctuation attached to a highlighted word should travel with the accent; leave it off when you want only the word itself colored. You can also highlight a single word by hand in the editor.

Highlight is a plan, not a guess at run time. Run Highlight with Generate so you can review the words it chose. Ignite then places the accents exactly as you approved them.

This documentation is indexed in plain text at /llms.txt.

Also worth reading

  • Furnace: rough cuts before captions Turn a long recording into a usable rough cut: trim pauses, clean up takes, let Cast choose short-form clips, and align an approved script.
  • Styles and Sparks Caption Styles, typography, colour and the animated details that make a subtitle look deliberate.
  • Ignite: captions and polish in one pass One pass places the captions and the polish around them: zooms, transitions, music, sounds, Flares, B-roll, Hide Cuts and a LUT.