> ## Documentation Index
> Fetch the complete documentation index at: https://help.descript.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Descript Terminology and Concepts

> Defines what Descript-specific terms mean and how to disambiguate overloaded words like clip, track, scene, project, drive, and publish. Use to interpret user language correctly in Descript context when answering questions.

Interpret the terms below in Descript context, not general usage. Use this to
understand what a user means — especially when they use an ambiguous, colloquial,
or legacy term — before answering.

## Account hierarchy (drive > workspace > project)

* **drive** — Central storage for a team's projects, layout packs, recordings, and AI Speakers. The billing and collaboration boundary. NOT a physical hard drive.
* **workspace** — A work area within a drive. Not the same as a project or a drive.
* **folder** — Organizes projects within the drive view. Exists at the drive level.
* **project file** — Imported audio/visuals plus Descript recordings. Source media, not the project container.
* **media library / media panel** — The in-project browser for shared drive files and stock media.
* **stock media** — Built-in cloud library of images, video, music, and SFX. Not user-uploaded.

## Collaboration & roles

* **drive member** — A user with drive access. Roles: Owner, Editor, Basic member.
* **viewer member** — A free seat with limited capabilities on a paid plan.
* **Editor** — A paid role with full editing rights.
* **project access** — Permission level for a specific project. Separate from drive membership.
* **project Editor / Editor collaborator** - A user with access to edit a specific project. NOT a drive Editor. Cannot use features that require media minutes or AI credits.

## Project hierarchy (project > composition > scene > script track OR layer > clip)

* **project** — The container holding compositions, files, and sequences. Top-level unit. A single project can hold multiple compositions.
* **composition** — An editable timeline inside a project for arranging and editing content. One project can have many compositions. Closest to a "sequence" in other media editing programs.
* **scene** — A segment created with the "/" key or "+" button that divides a composition into non-overlapping time segments and owns its layouts/effects. Like slides. NOT a "clip" or "take." May be called "cards" internally.
* **timeline** — An editor panel showing layers horizontally over time within a composition. For fine timing, wordbar adjustments, keyframes, cross-fades, and transitions.
* **layer** — A container for visual/audio elements (video, images, text, music) that stack. NOT part of the script track. Can contain multiple clips.
* **clip (technical)** — A visible/audible section of media within a layer. Only in audio/video layers. Used when discussing timeline editing, trimming, or splitting.
* **clip / clips (colloquial)** — Short videos from the Create Clips feature, for social. When a user talks about output quality, sharing, or quantity ("the clips were terrible," "send me the clips"), they mean Create Clips output.
* **sequence** — A container keeping grouped tracks in sync (e.g., separate video/audio from one recording); sometimes called a "multi-track sequence." NOT a composition.
* **track** — Individual rows within a sequence. In feedback, users sometimes mean a layer. Descript's docs use "tracks" in sequences and "layers" in compositions.

## Script & text

* **script / script track** — The foundational editing element containing transcribed media that keeps visuals, audio, and scenes aligned and enables text-based editing. NOT a screenplay.
* **transcript** — Text generated from audio, automatic or imported.
* **wordbar** — The timeline area showing words from the script's audio. Adjust word boundaries and gaps here.
* **word gap** — Pauses/silence between words. Use Shorten Word Gaps to reduce.
* **word boundary** — The start/end points aligning transcription to media. Manually adjustable.
* **filler words** — "um," "uh," "like," "you know." Detectable and removable via Remove Filler Words.
* **ignore** — Strikethrough text that non-destructively removes media from playback. NOT delete — the struck text stays visible in the script and is easily restored.
* **Correct / Correct mode** — Fixes the transcript text on screen without changing the audio or video underneath. "Fix the transcript" = Correct; "fix what was said / change the words" = Regenerate.
* **Write mode** — A mode for drafting script text before any audio exists; the first step for text-to-speech.
* **Edit for Clarity** — AI tool that trims rambling, repetition, and off-topic content, with an optional AI-speech reword. Lives in the "Sound Good" tools.
* **Remove Retakes** — AI that detects re-recorded phrases and marks the earlier takes for removal (as ignored text).
* **Smooth jump cuts** — AI (runs on Regenerate) that heals audio/video at jump cuts left by edits.

## AI, voice & generation

* **AI Speaker / AI voice** — A synthetic voice (stock or custom) that generates speech from script text. Current term for the voice itself.
* **text-to-speech (TTS)** — Generating spoken audio from script text. TTS and AI Speaker both current: TTS is the process; an AI Speaker is the voice it uses.
* **Overdub** — Ambiguous term with three possible meanings; disambiguate by intent. (1) Descript's legacy feature name, officially rebranded to **Regenerate** (fixing or changing recorded words). (2) Old Overdub also generated speech from typed text in a cloned voice — now **AI Speakers / text-to-speech**. (3) The traditional audio/video sense: layering a new recording on top of existing audio — in Descript, **adding an audio layer / recording into the timeline** (not an AI feature at all).
* **Regenerate** — AI feature that resynthesizes a speaker's recorded audio — and optionally their video (the "Audio & video" mode, sometimes called Video Regenerate) — to fix pronunciation, change words, heal jump cuts, or smooth pacing. Changing words requires an authorized custom AI Speaker and is English-only. NOT re-recording.
* **lip sync** — AI feature that resynchronizes lip movement to regenerated or translated audio. Noun: "lip sync." Verb: "lip-sync."
* **Speaker** — The combination of color + label (+ optional voice + optional avatar); can be saved to the drive for reuse.
* **avatar** — An AI presenter that speaks your script on camera, assigned to a Speaker. Users may say "AI presenter," "digital human," or "talking head." NOT a recorded person; NOT Eye Contact.
* **generate image / generate video** — Text-prompt creation of new images or video placed into a scene ("AI B-roll," "text to video"). NOT Regenerate (which fixes a speaker's own footage).
* **generate music / sound effects** — Prompt-based original background music and SFX.
* **Edit with AI** — AI image editing: add, change, or remove part of an image (inpainting-style). NOT Edit for Clarity (a script tool).
* **AI Video Maker** — Generates a whole editable video (script, voiceover, visuals, optional avatar) from a prompt, script, or footage; lives on the Home tab of the drive view. Also exists as a ChatGPT GPT, "AI Video Maker by Descript." NOT Quick design.
* **Quick design** — Turns an existing single-scene script project into a scene-based rough cut (scenes, layouts, optional B-roll). Needs media already in the script track. NOT AI Video Maker, which starts from nothing.
* **Underlord** — Descript's AI co-editor: performs edits, answers questions, suggests improvements. Users sometimes call it "Overlord" — they mean Underlord.
* **Studio Sound** — AI voice enhancement that reduces noise and echo. Has intensity settings. NOT a recording studio.
* **Eye Contact** — AI feature that adjusts gaze so the speaker appears to look at the camera.
* **Green Screen** — AI background removal. No physical green screen needed; uses AI, not chroma key.
* **Create Clips** — AI tool that finds highlights and creates new compositions as short-form social videos. When users say "clips" about output/quality/sharing, they mean this feature's output.

## Recording

* **Descript Rooms / Rooms** — Real-time, multi-participant remote recording. NOT generic meeting rooms.
* **SquadCast** — The precursor to Descript Rooms and now not in active development. Legacy SquadCast plans still work, and billing consolidates into Descript once accounts are linked. Descript Rooms is its successor for remote recording.
* **Control Room** — The producer's "backstage" view and controls during a Rooms session, where Producers monitor and coordinate without being seen or recorded.
* **screen recorder** — Available only via the Descript desktop app. Also available as a standalone OS-toolbar recorder for quick captures without opening the main app.
* **editor recorder** — The in-app multitrack recorder for audio, video, and screen.
* **teleprompter** — Built-in feature that displays script text while recording.

## Visual editing

* **Scene Editor** — The official name for the panel where you arrange visuals in a scene. Many users call it the **canvas** — the common term in other editors, and still used informally in Descript for the visual surface. Not shown for audio-only projects.
* **layout / layout pack** — A collection of pre-made scene designs (arrangements, text styles, positioning); can be reused for repeatable content across projects, like intros and outros. Formerly called a **template**, and many users still say "template" for this — disambiguate (see below).
* **template (Underlord template)** — A saved, pre-planned Underlord workflow/prompt for making repeatable edits. "Template" used to mean a layout/layout pack, and users often still use the term template. Disambiguate by intent, and direct users to layouts, saved scenes for reusable visual design or content.
* **animation** — A change to layer properties between keyframes. Presets or custom.
* **keyframe** — Defines properties at an animation's start/end and its duration.
* **transition** — A visual effect between scenes. "Smart Transitions" are AI-powered.
* **gap clip** — A silent/blank clip used for timeline spacing or pauses. Technical term — unrelated to Create Clips.
* **Automatic Multicam** — AI that cuts between camera angles across a multi-track recording based on who's speaking. NOT Center active speaker.
* **Center active speaker (beta)** — Auto-reframes a single combined video+audio clip to keep the talker centered. Does not work with multi-track sequences. NOT Automatic Multicam.
* **blur (four different features)** — **Blur**: softens the layer it's applied to. **Glass blur**: a masking layer that blurs whatever shows through it (for hiding a face, name, or number). **Blur speaker background**: depth-of-field blur behind the subject. **Zoom blur**: radial motion blur for energy. If a user just says "blur," clarify which.

## Export & publishing

* **share page / Descript web link** — Exporting to a "Descript web link" (or just "web link") — also called publishing to a share page — creates a standalone page on share.descript.com with its own URL, where anyone can watch or listen in a browser (no account or download). "Share page," "web link," "Descript link," and "share link" all mean this. NOT a document page; NOT a local file export.
* **timeline export** — EDL/XML/AAF for other editors.
* **publish** — Put a composition online. Historically this meant exclusively creating a Descript web page (share page). Now it's one destination inside the **Export** flow — "export to Descript (web link)" — alongside exporting to hosting platforms (YouTube, podcast hosts). So "publish" and "export to a web link" now mean the same thing. Still distinct from **local export** (downloading a file to your computer).
* **replace file** — Loads a new file in place of an existing one and deletes the replaced file. Destructive.

## Repurpose & translation

* **Repurpose** — The AI-tools category for turning content into other formats or languages (Translate, Dub, etc.).
* **Translate** — Generates translated captions and subtitles.
* **Dub / Dub speech** — Generates a spoken translated voiceover. Pairs with lip sync to match mouth movement.

## Brand settings

* **Brand Studio** — The brand-settings suite: fonts, colors, layout packs, glossary. Asset management, not a design tool.
* **transcription glossary** — A custom word bank that improves transcription accuracy for names and technical terms. NOT a list of definitions.
* **do not translate (DNT) list** — Terms that stay untranslated when dubbing (brand names, technical terms).

## Common misinterpretations

* **"clip" / "clips"** — Output/quality/sharing/quantity context ("the clips were terrible," "send me the clips") = Create Clips short videos. Timeline/editing context ("split the clip," "trim the clip") = a technical media segment in a layer.
* **"the scene"** — A scene boundary/segment. NOT a shot or take.
* **"the script" / "my script"** — The transcription/script track. NOT a written screenplay.
* **"the project"** — The Descript project container. NOT a single video file.
* **"my composition"** — A container within a project with its own script track/sequence and layers. NOT the entire project.
* **"my drive"** — A Descript drive. NOT a hard drive.
* **"the track"** — A layer, timeline, or sequence track. Clarify if ambiguous.
* **"the canvas"** — The Scene Editor / the visual surface. NOT a separate feature.
* **"regenerate"** — AI audio/video fix or word change. NOT re-recording.
* **"overdub"** — Ambiguous. Fixing/changing a recorded word → Regenerate. Typing text to generate speech in a cloned voice → AI Speaker / text-to-speech. Traditional A/V sense of layering a new recording over existing audio → adding an audio layer / recording into the timeline.
* **"blur"** — Could be Blur, Glass blur, Blur speaker background, or Zoom blur. Clarify which.
* **"fix the transcript"** — Correct (text only). "Fix what I said / change the words" — Regenerate (new audio).
* **"make me a video" / "AI video"** — AI Video Maker (builds from scratch) vs Quick design (designs an existing project). Clarify which.
* **"publish"** — Put a composition online — now framed as "export to Descript (web link)." Distinct from a local export (downloading a file).
* **"the link" / "web link" / "share link" / "Descript link"** — The Descript web link / share page on share.descript.com. NOT a local export file.
* **"overlord"** — The user means Underlord.
* **"backstage"** — The Control Room in Descript Rooms (SquadCast's term for the same producer space).
* **"template"** — Officially, Underlord templates are a saved workflow/prompt. But many users looking for a "template" mean a **layout / layout pack** (saved scenes to be used for repeatable visual design or content), which was formerly called a template. Disambiguate by intent.

## Disambiguation rules

* If "clip/clips" appears: output/quality/sharing/quantity context = Create Clips short videos; timeline/editing/trimming context = a technical media segment.
* The script track is special: it is both the transcription and the foundational backbone of a composition.
* Scenes ≠ clips: scenes are structural dividers; clips are media segments OR short-form outputs depending on context.
* Generate vs Regenerate: "generate" image/video/music = create NEW AI media; Regenerate = fix a speaker's EXISTING footage.
* Correct vs Regenerate: Correct changes only the transcript text; Regenerate changes the actual audio/video.
* Edit with AI vs Edit for Clarity: Edit with AI works on images; Edit for Clarity works on the script.
* AI Video Maker vs Quick design: AI Video Maker builds a video from a prompt; Quick design designs a project that already has media in the script track.
* Automatic Multicam vs Center active speaker: multi-track camera switching vs single-clip reframing.
* Four blurs: Blur (own layer), Glass blur (what shows through a mask), Blur speaker background (behind the subject), Zoom blur (motion). Clarify which.
* AI terms evolve: Overdub split into Regenerate (fix recorded words) and AI Speakers / text-to-speech (generate speech from typed text); a user with an A/V background may instead mean the traditional sense — layering a new recording over existing audio (add an audio layer). Officially rebranded to Regenerate; map by the user's intent. Prefer current terminology.
* Capitalization signals a branded feature: if a term is capitalized, it's likely a specific product feature.