DocsGet started

Projects

Captions

On the phone

In the video editor, tap Captions, then Generate captions. Check the language and pick a style, tap Generate captions, and every word lands on the video at the moment it is said.

The Generate captions sheet

ChoiceWhat it decides
Generate fromThe video's own speech, or one voiceover or song on the audio lane. A sound with words in it is captioned on its own timing.
Spoken languageAuto-detect unless you pick one. It listens on this phone first, and the audio stays on the phone when you speak its language or another language it hears. A language marked Cloud, uses AI, or one auto-detect cannot place, is sent to OpenAI's Whisper through MORR once AI is on in Settings > AI; MORR keeps the words, not the audio. The phone's own list depends on the iPhone and its iOS; the cloud adds the rest of Whisper's 57 languages. Auto edit and other one-tap passes never send audio to the cloud on auto-detect.
PositionClear of faces to begin with: it looks at the take and moves each line off the face and lap. Or top, middle, lower third, or the style's own.
Transcription controlsA page of its own: remove filler words, a profanity filter (fk or **), punctuation on or off, casing (as spoken, ALL CAPS, all lowercase, sentence case), words per line, lines, the shortest time a caption stays up (so a quick word is still on screen long enough to read; 0.8 s is what TV captions use), and a timing nudge for every caption in tenths of a second. A line of your own captions shows the result as you change them.
StyleEvery style moving on your own first line, filed in groups with a chip per group to jump to it. A first pass starts in Soft serif: lowercase serif type in a soft white, four words a line, low in the frame and clear of the face.

A bar shows the listening with Cancel; close the sheet and keep editing, and the captions land when it is done. If it cannot finish, it says why and offers the next step: Open Settings when speech recognition is off, Hear it in the cloud when the phone cannot hear that language, Allow AI and try again when cloud captions are off, Pick another source when no one talks. Generating again replaces every caption, fixes included, and one undo puts them back.

What it hears on the phone

MORR listens to the whole take, so no word is lost on a join. Sentences start with a capital and a page ends where the sentence does; a one-word page never shows a stray comma. Numbers read as one word (38%). Your name, brand and catalog products are spelled your way. A language the phone can hear stays on the phone. For one it cannot hear, the audio goes to OpenAI, and only after you allow AI.

If the speech sounds like a language the phone cannot hear, Captions says so once (This sounds like Spanish) and Choose language opens the sheet. When a few words are uncertain, a line says 3 words to check, and Review opens the transcript on only those words. Nothing shows when every word is sure.

Fixing the words

ToDo this
See every lineTranscript in the Captions panel. Each line has its time; tap the time to play from there.
Find words to checkWords the recognizer was less than half sure of are underlined in amber. To check shows only those lines; It's right clears the mark.
Fix a lineTap it and change the words. The same number of words keeps their timing.
Split, join, moveIn an open line, tap a word to start a new line with it. Join with next, 0.1 s earlier, 0.1 s later and Delete line sit under it. Tapping a caption block opens that line.
Fix a name everywhereFind and replace changes every caption at once, whole words only, keeping a capital.
Trim or move oneCaptions sit on blue rows above the text items, one block per caption. Drag a block or its edges; they snap like anything on the timeline, and a word trimmed out of the block leaves the caption.
Edit the video under themChange a clip's speed, reorder clips, trim one or delete one, and MORR's own captions are written again from the words it already heard, so every word stays on its sound, and one undo takes both back. Captions you typed or fixed by hand stay where they are.
Change the controls laterTranscription controls in the Captions panel changes casing, punctuation, swears, lines and timing on every caption at once, as one undo, without hearing the take again.

The 102 styles

GroupStylesWhat they are for
Trending32What is on the feed this month, and the first group in the picker: big pink type stacked over the chest, a blue or lemon chip that jumps to the word being said, a white pill around the line, tall gold capitals, a retro peach script, a typed label on paper, Word swap (one word at a time, each in its own face and color), Top clean (a lowercase sans word at a time high in the frame, the key word in italic serif), Hand note (small hand lettering at the top, the key word bigger), Chip stack (each word on its own crooked white card, the last one white on red), Fade pill (a white pill that reads along, each word turning dark as it is said), Italic punch (the loud word in a heavy italic), Navy serif (a navy pill, a thin serif), Search bar (the line in a thin white ring), Circled (a thin ring on each word as it is said, one ring at a time), Two voices (a light sans, the key word in a cream italic serif), and three that set the words behind your head. It changes as the feed does.
Word stack5Each word placed and sized on its own in two or three faces: small words small, the main word big, building up word by word clear of your face. Editorial mix leads.
Clean6White type with a soft shadow, for anything. Soft serif, the style a new video starts in, leads it and the whole picker.
Bold6Heavy capitals with an outline, one key word in color.
Karaoke highlight5The line stays up and the spoken word fills with color.
Word pop5One word at a time, big, popping in.
Boxed5Type on a plate, a block per word, or a highlighter on the spoken word.
Outline4Outlined letters, including Hollow, where only the outline shows.
Neon glow4Pale type with a colored glow.
Handwritten4Pen, marker and chalk, including a note on lined paper.
Editorial serif5Serif type from columns and covers.
Retro6Typewriter, date stamp, label tape and a seventies script.
Y2K4Chrome, bubblegum, pixel type and glitter.
Soft4Lowercase on pastel plates.
Minimal subtitle4Small, low and still, like a film's subtitles.
Commentary3Small words at chest height that change as you talk, for breakdowns.

Words arrive by pop, bounce, slide up, typewriter, fade or karaoke fill, depending on the style. Key words sets one word per line apart: colored, bigger, or both, in any color or your brand's. With no caption picked, a style goes on the whole lane; with one picked, Style goes on chooses every caption or only that line. Every style is drawn the same in the preview and the saved video, at 9:16, 4:5 and 1:1; a style that would run off a shorter frame shrinks to fit.

My styles

  1. 1Start from a stylePick the closest one, then Make it yours in the Captions panel.
  2. 2Change itFace, from the caption faces or the whole font library through the same picker the text tools use (shelves, search, favorites and your own fonts), weight, case, size, text color, outline (its color and width), shadow, box, the spoken word's color and how words arrive. Every color row ends in a color wheel for any color. The preview moves as you change it.
  3. 3Save itName it and tap Save, which stays at the foot of the sheet however far you scroll. If the phone could not keep it, the sheet stays open and says so, with Save again. It goes on every caption or only the picked line, and waits under My styles at the front of the picker. Saved on this phone; each video keeps its own copy, so deleting a style leaves finished videos as they are.

Ask in the editor reaches all of this: “captions in Spanish”, “make the captions word stack”, “highlight key words in yellow”, “remove the ums from captions”, “one word at a time”. Captions do not yet tell two speakers apart: neither the phone's recognizer nor Whisper marks who is talking.