Underlord
a chat-based AI co-editor that edits the video for you
A chat-based AI agent inside the editor that carries out editing requests: cutting, captions, clips, animations, translation, audio effects, turning slides or documents into a narrated video, and searching the web for current facts. It runs on a large language model the user can choose from several providers, uses the AI tools on the user’s behalf, and Descript warns that it can get things wrong. Saved instructions can be reused as skills.
Where to turn off Underlord: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Automatic transcription
turns every recording into editable text with third-party AI models
Every audio or video file added to a project is transcribed automatically, so the recording can be edited by editing its text, captioned and searched. Descript uses third-party AI transcription models and picks one by default; a user can choose another model and add a glossary of names and terms.
Where to turn off Automatic transcription: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Automatic speaker detection
AI that tells the voices in a recording apart and labels each speaker
When a file is imported and transcribed, Descript’s AI analyses the audio, finds the distinct voices and asks the user to name each one from a voice sample; the names are then applied throughout the transcript. It works on files up to 5 hours long.
Where to turn off Automatic speaker detection: see below.
Something wrong here? Tell us ↓
Sound good AI tools
AI clean-up of speech: Edit For Clarity, Studio Sound, Remove Filler Words, Remove Retakes, Shorten Word Gaps, Add Chapters
A group of one-click AI tools for spoken audio: Edit For Clarity trims rambling, repeated and off-topic parts of the script and can generate new AI speech to reword small sections; Studio Sound removes background noise and echo; Remove Filler Words and Remove Retakes take out ums and repeated takes; Shorten Word Gaps tightens long pauses; Add Chapters generates chapter markers and titles. Removed words are kept as crossed-out text the user can review.
Where to turn off Sound good AI tools: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Eye Contact
AI that redirects a speaker’s eyes so they appear to look into the camera
Changes the eyes of the person on screen with AI so that a take recorded while reading off-screen looks as if they were looking straight into the camera. It works on one clear face in the frame and is applied as an effect on a video layer.
Where to turn off Eye Contact: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Look good AI tools
AI design and framing: Quick Design, Green Screen, Center Active Speaker, Automatic Multicam, Portrait Lighting, Uplighting
A group of AI tools for the picture: Quick Design builds a first draft of a video with layouts and AI-generated or stock B-roll; Green Screen uses AI to find and remove the background; Center Active Speaker reframes the shot around whoever is talking; Automatic Multicam switches between camera angles based on who speaks. The Portrait Lighting and Uplighting effects also use AI.
Where to turn off Look good AI tools: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Regenerate
AI that rewrites what a speaker said and regenerates their voice and face to match
Uses AI to regenerate parts of a recording without re-recording: smoothing jump cuts and awkward pacing, removing a distracting noise, or changing the words a person said by generating new audio in their authorised voice clone. In video it can regenerate the speaker’s face as well as the audio, and Descript asks for consent the first time video is regenerated because it changes the speaker’s performance. English only.
Where to turn off Regenerate: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
AI speech
text-to-speech voiceover from a typed script, in stock AI voices
Generates spoken audio from a typed script in a stock AI voice or the user’s own voice clone, in about 20 languages, with tone directions such as whisper or laugh. A recorded voice can also be replaced by an AI voice reading the same script. Descript uses speech models from ElevenLabs.
Where to turn off AI speech: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
AI Speakers (voice clones)
a clone of a person’s voice, made after they record a consent statement
Creates a custom AI Speaker that speaks in a real person’s voice, so new lines can be generated in that voice by typing. The person must record a consent statement in English, in the app or as a recording sent to the user; Descript says it will not clone a deceased or non-consenting person or audio from an artificial source. AI Speakers can be shared from the Drive view.
Where to turn off AI Speakers (voice clones): see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Translate and dub
AI translation and dubbing: Translate captions, Dub speech, Lip sync
Translate captions turns a finished video’s captions and transcript into other languages; Dub speech replaces the original audio with a translated AI voice; Lip sync changes the speaker’s mouth movements to match the translated audio. Each translation creates a new composition.
Where to turn off Translate and dub: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Repurpose AI tools
AI that picks the best moments of a long video: Create clips, Create highlight reel, Find highlights
AI tools that scan a long video or podcast and pick out its most engaging or important moments: Create clips turns them into short clips, Create highlight reel joins them into a teaser or recap, and Find highlights marks them in the script. The user sets the number and length of clips and can give a topic to guide the choice.
Where to turn off Repurpose AI tools: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Write and Publish AI tools
AI-written text: Brainstorm, Write a script, Write an outline, Rewrite, Draft a title, Summarize, Show notes, YouTube description, Draft a social post, Draft a blog post
AI tools that write text from a prompt or from the video’s transcript. Before recording: Brainstorm, Write a script, Write an outline and Rewrite. After editing: Draft a title, Summarize, Show notes (with chapters and timestamps), YouTube description, Draft a social post and Draft a blog post. They work on the whole script or a selected part.
Where to turn off Write and Publish AI tools: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Generate media
AI images, video, music and sound effects from a prompt, plus Edit images with AI, Extend video, Start and End Frame and YouTube thumbnails
Generates images and short video clips from a text prompt, straight into a scene, using models from other companies such as Google, OpenAI, Black Forest Labs, Kling, MiniMax and PixVerse. Related tools: Edit images with AI changes part of an image from a prompt, Extend video adds frames to a clip that is too short, Start and End Frame creates a transition between two images, Generate a YouTube thumbnail, and Generate AI music and sound effects.
Where to turn off Generate media: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
Avatars
an AI presenter that speaks the script on camera, made from a prompt, a photo or a stock gallery
Adds an AI presenter who speaks the script on camera, lip-syncing to recorded or AI-generated audio, so no one has to be filmed. The avatar is generated from a text prompt, from an uploaded photo, or chosen from Descript’s stock gallery, and is then attached to a speaker in the script. Avatars are generated with Kling Avatar models.
Where to turn off Avatars: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
AI Video Maker
builds a whole video — script, AI voiceover, visuals and an optional avatar — from a prompt
Generates a complete, editable video from a prompt, a pasted script or existing footage: the script, a stock AI voiceover, AI-generated or stock visuals in a chosen style, and an optional avatar. It is started from the Home tab of the Drive view, and also from ChatGPT through Descript’s GPT.
Where to turn off AI Video Maker: see below.
- Where to check the switch
- No switch found — look in your Descript settings or ask your administrator
Something wrong here? Tell us ↓
