Descript Review: Text Editing, Underlord, Rooms, API and MCP

Descript's June 2026 release focuses less on generating a complete artificial video and more on repairing the footage creators already have. Video Regenerate can synthesize a natural bridge across a jump cut created when filler words, retakes, or pauses are removed from the transcript.

Portrait Lighting and Uplighting address poor recording conditions by separating the speaker from the background or brightening dark footage. New visual effects add stylized treatments, while Edit with AI lets users circle an unwanted region in an image and remove or replace it without opening a separate photo editor.

Descript also added a YouTube thumbnail generator that can use a frame from the project, a chosen style, and a title to produce a designed thumbnail. These tools sit alongside Studio Sound, Eye Contact, Green Screen, Regenerate Speech, and Underlord.

The product direction is important because it strengthens Descript's original buying argument. The platform is not trying only to replace the camera with generation. It is using AI to make ordinary recorded video easier to fix, edit, repurpose, and publish.

WhatAI Decision Box

Descript now combines transcript-based editing with Underlord, remote Rooms recording, Automatic Multicam, AI video enhancement, generated media, publishing, API automation, and MCP connections. Its strongest value is turning spoken-word production into one connected workflow rather than one isolated editing feature.

Descript treats the transcript as the main editing surface. Deleting words removes the corresponding media, while AI tools can clean audio, remove retakes, smooth jump cuts, correct speech, generate clips, add captions, and create supporting visuals. Rooms and screen recording also move capture into the same production environment.

Fast transcript editing can hide visual continuity problems, incorrect transcriptions, weak cuts, or generated media that does not fit the story. Underlord can perform many actions, but its output needs review and its credit cost varies. Advanced cinematic finishing, compositing, sound design, and deterministic motion work still suit specialist software better.

Use Cases

Pricing varies by plan and region — see current pricing.

Plan features change — last updated: 2026-07-22.

Descript Community Discussions

Explore community discussions. Ask and answer questions on Descript to grow and learn together.

Descript Pros & Cons

How to Get Results with Descript: Step-by-Step Workflow

1. Define the final asset

   Specify the audience, channel, duration, aspect ratio, recording format, source rights, deadline, and evidence that will make the podcast or video ready to publish.

   Decision point: Use Descript when spoken words drive the edit. Use a specialist editor when visual timing, compositing, or motion design drives the project.

2. Choose the recording source

   Record locally, capture the screen, use Rooms for separate remote tracks, import a Zoom cloud recording, or upload existing media.

   Decision point: Use Rooms when separate participant video tracks and Automatic Multicam matter. Use Zoom import when the meeting platform cannot change.

3. Check the transcript first

   Correct names, terminology, speaker labels, numbers, and important quotations before using the transcript as the editing surface.

4. Create the structural edit

   Remove irrelevant sections, repeated takes, filler words, long pauses, and digressions by editing the transcript. Preserve a recoverable version.

5. Use Underlord for bounded tasks

   Ask Underlord to perform one clear job such as selecting strong takes, creating a short clip, improving clarity, or building scenes.

   Decision point: Use manual controls when the edit depends on subtle pacing, legal wording, or precise visual continuity.

6. Repair sound and speech

   Apply Studio Sound selectively, verify noise removal, and use Regenerate Speech only for approved corrections in an authorized voice.

7. Inspect every transcript cut

   Watch the boundaries created by deletions. Use Video Regenerate, B-roll, a layout change, or a manual cut when movement visibly jumps.

8. Build the visual layer

   Apply scenes, layouts, Smart Transitions, screen recordings, B-roll, captions, images, generated media, and brand assets.

9. Create derivative assets

   Generate social clips, show notes, summaries, thumbnails, captions, descriptions, and translated versions from the approved master.

10. Review generated and translated content

    Check facts, quotations, pronunciation, translation, captions, avatar output, visual continuity, licensing, and accessibility.

11. Publish or automate

    Export manually, publish a Descript web page, send to a supported platform, or use API, Zapier, CLI, or MCP for repeatable workflows.

    Decision point: Remember that the API cannot currently export finished media files directly.

12. Measure usage and rework

    Track media minutes, AI credits, top-ups, editor seats, transcript corrections, rejected AI edits, review rounds, and time to approved publication.

Descript Gotchas and Limits to Know Before You Start

• Free includes sixty media minutes per month but only one hundred one-time AI credits
• Unused monthly media minutes do not roll over
• Unused monthly AI credits do not roll over
• Creator and Business can purchase top-ups while lower plans have fewer usage options
• Each paid Editor seat contributes to and charges for the shared drive resource pool
• Underlord messages consume AI credits even when the user is only chatting
• Underlord can charge for both agent reasoning and each tool it applies
• Generated media costs depend on the selected model and output
• Studio Sound and Remove Filler Words have capped per-file credit costs but still consume credits
• Transcription can misidentify names speakers numbers and specialist language
• Removing words can create visible jump cuts and unnatural speech timing
• Video Regenerate can smooth a cut but may introduce synthetic facial or motion artifacts
• Regenerate Speech should only use authorized voices and can sound inconsistent across long passages
• Automatic Multicam requires separate video feeds from Rooms for its full workflow
• Zoom imports provide separate audio tracks when configured but not separate participant video files
• The API can publish web links but cannot directly export finished media files
• API tokens are drive scoped and anyone holding one can consume usage
• YouTube URLs are not supported as import sources through the current API
• Brand Studio custom avatars and stronger governance require Business or Enterprise
• Descript remains cloud dependent and is not a fully offline editor

Which Descript Feature Fits Your Use Case

How Well Descript Fits Common Use Cases

Starter Prompts for Descript

Create a conservative rough cut of this interview. Remove false starts, duplicated answers, long off-topic sections, and filler words that do not affect meaning. Preserve every important claim and quotation. Mark uncertain cuts for review instead of deleting them.

Review this talking-head video for poor audio, visible jump cuts, weak lighting, eye contact, empty pauses, and distracting backgrounds. Propose the lowest-risk fix for each issue before applying anything. Do not generate or replace the speaker's words without approval.

Identify three self-contained moments from this approved episode for LinkedIn, YouTube Shorts, and Instagram Reels. Each clip must preserve the speaker's meaning, open with a clear hook, include accurate captions, and end without cutting off necessary context.

Design a Descript API workflow that imports a recording URL, creates a project in a named folder, applies Studio Sound and filler-word removal, generates a thirty-second highlight, publishes a Descript web link, and records media-minute and AI-credit usage. Include job polling, retries, token security, errors, and the manual step required for final file export.

Translate and dub this approved training video into Indonesian. Keep product names in English, preserve mandatory policy language, generate captions, and flag terminology or timing that requires native review. Do not change the instructional sequence.

Prompt pattern: Using [recording or project], create [approved output] for [audience and channel]. Preserve [meaning claims words and rights]. Apply [specific edits]. Flag [uncertainty and artifacts]. Verify [transcript audio visual continuity credits and publication requirements].

Iteration tip: Approve the structural transcript edit before adding enhancement, generated media, clips, translation, or synthetic speech.

Descript remains one of the clearest productivity gains in spoken-word media because the transcript is both a document and an edit decision list. A podcaster, educator, marketer, or product team can record, transcribe, remove mistakes, clean sound, add visuals, create clips, publish, and collaborate without mastering a traditional timeline first.

The product has expanded around that foundation. Underlord can inspect a project and perform multi-step edits. Rooms records remote guests locally in separate tracks, while Automatic Multicam chooses layouts according to the active speaker. Recent video-enhancement tools can smooth jump cuts, improve poor lighting, clean images, and create YouTube thumbnails. Descript also exposes imports, Underlord edits, project management, and web publishing through an API, Zapier integration, command line, and MCP connections.

Free is useful for evaluation but includes only sixty monthly media minutes and one hundred one-time AI credits. Hobbyist suits occasional individual work. Creator is the practical professional plan because it adds thirty media hours, eight hundred monthly AI credits, 4K export, premium generation, stock media, and top-ups. Business adds Brand Studio, translation, dubbing, custom avatars, more credits, and team support. Enterprise is for organizations that need SSO, SCIM, audit logs, custom retention, AI controls, legal terms, and dedicated support.

Choose Descript over Riverside when editing and repurposing are more important than recording alone. Choose Adobe Premiere Pro, Final Cut Pro, or DaVinci Resolve when advanced visual and audio control matter more than transcript speed. Choose ElevenLabs when voice generation, dubbing, or conversational audio infrastructure is the primary need.

Evaluate Descript using one real episode or business video. Record multiple speakers, remove mistakes, apply Studio Sound, create a clip, smooth one jump cut, correct one spoken line, publish through the intended channel, and measure media minutes, AI credits, correction time, transcript errors, and the amount of specialist editing still required.

What is Descript best used for in 2026?

Descript is best used for spoken-word audio and video production, including podcasts, interviews, tutorials, screen recordings, social clips, marketing videos, training, and internal communication.

How much does Descript cost?

Descript has a Free plan, Hobbyist at twenty-four dollars monthly, Creator at thirty-five dollars monthly, Business at sixty-five dollars per person monthly, and custom Enterprise pricing. Annual billing reduces the effective price.

What are Descript media minutes?

Media minutes measure uploaded or recorded audio, video, Rooms sessions, screen recordings, and other media. Unused monthly media minutes do not roll over.

What are Descript AI credits?

AI credits pay for features such as Underlord, Studio Sound, filler-word removal, Green Screen, Eye Contact, AI Speech, avatars, and generated video. Unused monthly AI credits do not roll over.

What is Underlord?

Underlord is Descript's agentic editing assistant. It can understand a project, answer questions, apply edits, clean audio, remove mistakes, generate scenes and clips, write supporting copy, and use other Descript tools.

What are Descript Rooms?

Rooms is Descript's remote recording environment. It can record participants locally in separate high-quality tracks, support video up to 4K, recover interrupted guest uploads, and send the recording directly into a project.

What is Automatic Multicam?

Automatic Multicam uses separate Rooms video tracks to switch scenes and layouts according to who is speaking. It reduces the manual camera switching required for interviews and video podcasts.

Can Descript fix jump cuts and bad lighting?

Yes. Descript's newer enhancement tools include Video Regenerate for smoothing jump cuts, Portrait Lighting, Uplighting, Studio Sound, visual effects, and AI-assisted image cleanup.

Does Descript have an API?

Yes. The current API can import media, trigger Underlord edits, manage projects, track jobs, and publish to Descript web links. Direct finished-media file export is not currently supported through the API.

Can Descript connect to ChatGPT or Claude?

Yes. Descript provides an MCP server for ChatGPT, Claude, Cursor, Codex, and compatible assistants. API and MCP activity can consume the drive's media minutes and AI credits.

Related Audio & Voice Tools

Explore the Network