Affiliate disclosure: This article contains affiliate links. If you sign up for Descript through one of these links, I may earn a commission at no extra cost to you. I only recommend tools that I believe can help podcasters and content creators. This disclosure is placed here so you can see it before reaching any product links.

Recording a podcast or video can be the easy part.

Editing is where many creators get stuck.

You finish a strong interview, but now you need to remove mistakes, shorten long answers, clean up the audio, add captions, create social clips, write show notes, and prepare everything for YouTube.

That can turn a one-hour interview into several hours of work.

Descript is built to make this process easier. It lets you edit audio and video by changing the written transcript. Its AI assistant, Underlord, can also complete many editing and content-repurposing tasks from plain-language instructions.

Descript has released several useful features during 2026. These updates are not only about adding more AI. Many of them improve the actual outcome for creators.

They can help you produce:

  • Cleaner podcast audio
  • Tighter interviews
  • Smoother video cuts
  • Better-looking webcam recordings
  • Short videos for social media
  • YouTube thumbnails
  • Show notes and descriptions
  • Remote interviews with fewer setup problems
  • More content from every recording

If editing is stopping you from publishing consistently, these updates are worth exploring.

Try Descript and see how transcript-based podcast and video editing works.

What Is Descript?

Descript is an audio and video editing platform designed around text.

When you import a recording, Descript creates a transcript. You can then edit the recording by editing the words on the screen.

Delete a sentence from the transcript, and Descript removes that section from the audio or video. Copy and paste a section of text, and you can move that part of the recording.

This makes the editing process feel more like working in a document than using a traditional video timeline.

You can still access a timeline when you need detailed control. You are not locked into using only the transcript. The transcript gives beginners a more familiar place to start.

Descript includes tools for:

  • Podcast editing
  • Video editing
  • Remote recording
  • Screen recording
  • Automatic transcription
  • Captions and subtitles
  • Audio cleanup
  • AI speech
  • Social clips
  • Show notes
  • YouTube descriptions
  • AI-generated images and video
  • Publishing and exporting

The platform can be used for a simple audio podcast, a video interview, a YouTube tutorial, a presentation, or a short social video.

Why Text-Based Editing Matters

Traditional editing software can be intimidating.

You may see several tracks, waveforms, tools, keyframes, and menus before you even make your first cut. Learning what each control does takes time.

Text-based editing removes some of that friction.

You can read through your conversation and identify:

  • Repeated questions
  • Long introductions
  • Unnecessary side stories
  • False starts
  • Technical problems
  • Filler words
  • Strong quotes
  • Emotional moments
  • Possible short clips

This can be especially useful when editing a long podcast interview.

Instead of listening through the entire recording every time you want to find a specific answer, you can search the transcript for a name, topic, or phrase.

The transcript also gives you a quick view of how much each person spoke. You may notice that the host’s question is longer than the guest’s answer. You may find that the same point was discussed three times.

You still need to listen to your final edit. A written transcript cannot show every detail of a person’s tone, timing, or emotion.

The transcript makes it easier to shape the conversation before completing that final listening pass.

Underlord Is Becoming a Real AI Editing Assistant

One of Descript’s biggest areas of development is Underlord, its AI editing assistant.

Underlord works inside your project. You describe what you want, and it uses Descript’s editing tools to complete the request.

You could ask Underlord:

Remove false starts, repeated points, and filler words. Shorten pauses longer than two seconds, but keep emotional pauses and natural laughter.

You could also ask:

Edit this interview into a focused 30-minute episode. Keep the guest’s personal stories and practical advice. Remove technical setup, repeated explanations, and unrelated tangents.

For a YouTube video, you might ask:

Turn this recording into an eight-minute video for beginners. Start with the strongest question, keep the clearest examples, and remove anything that does not support the main topic.

These requests give Underlord a clear editing goal. You are not simply asking it to “make the video better.” You are telling it who the content is for, what should remain, and what should be removed.

Underlord can also help with:

  • Removing filler words
  • Removing repeated takes
  • Tightening word gaps
  • Improving audio
  • Creating clips
  • Adding captions
  • Finding specific moments
  • Creating a rough story structure
  • Writing show notes
  • Drafting YouTube descriptions
  • Developing titles
  • Generating promotional copy
  • Adding visuals
  • Creating a trailer

Descript says Underlord can read the script, review the video, suggest changes, and complete several editing steps from a single request. Learn more about Underlord from Descript.

Attach Specific Context to Your Editing Request

A useful Underlord update is the ability to attach specific parts of your project to a request.

You can reference a:

  • File
  • Scene
  • Layer
  • Timestamp
  • Time range
  • Section of the transcript

This helps when you do not want Underlord to work on the entire project.

For example, you might select a five-minute section and ask:

Tighten only this part of the interview. Keep the guest’s story about his father, but remove my repeated questions and the discussion about setting up the microphone.

This gives the AI a clearer target.

It can reduce the chance that the tool changes a different section or removes something you wanted to keep.

For complex podcast episodes, this is more practical than trying to explain where a specific conversation appears.

Underlord Chat History Is Saved

Underlord conversations can now remain connected to the project.

If you close the project and return later, your latest conversation can still be available. Older conversations can be reviewed, although Descript notes that older chats may be read-only.

This can help creators who edit an episode over several days.

You might use one conversation to clean up the interview, another to create clips, and another to write promotional content.

Saved chats also let you revisit prompts that produced good results. You can copy the structure of a successful request and use it for the next episode.

That can help you develop a consistent editing workflow instead of starting from scratch every week.

AI Reviews Underlord’s Changes

Descript has added an automatic review process after Underlord completes an editing turn.

A second pass checks the changes against the original request. It looks for problems such as:

  • Unintended deletions
  • Missed instructions
  • Edits applied to the wrong section
  • Overcorrections
  • Changes that went outside the requested scope

If the review detects a problem, Underlord can attempt to correct it.

This does not replace your review. It adds another check before you inspect the edit yourself.

This matters because an AI edit can be technically clean while still hurting the story. It might remove a pause that carries emotion or shorten an answer so much that the meaning changes.

The creator still makes the final call.

Descript’s own 2026 editing guidance says AI can handle tasks such as removing filler, tightening pauses, and finding usable takes, but the human still decides what the video is about. Read Descript’s AI editing guide.

Smoother Jump Cuts for Video Podcasts

One of Descript’s newest features focuses on smoothing jump cuts.

A jump cut happens when you remove part of a video and join the remaining pieces. The speaker’s head, hands, or body may suddenly move between frames.

A small cut might look like the person jumped forward.

Descript rebuilt its Regenerate process around these edits. When you delete words or select an existing cut, you can choose to regenerate:

  • The video
  • The audio
  • Both the video and audio

The goal is to make the transition feel less distracting.

This can be useful for video podcasts, tutorials, and talking-head videos. You can shorten a long answer without hiding every edit behind stock footage.

Smoother cuts can also make short videos look more professional. Viewers may not know why a cut feels distracting, but they can notice when a video feels rough.

You should review every regenerated section closely. Look for strange facial movement, lip-sync problems, changes in voice, or unnatural motion.

When the feature works well, it can save time. When a result looks artificial, you can use a different cut or cover it with relevant B-roll.

Edit for Clarity Without Removing the Person’s Personality

Descript’s Edit for Clarity tool can help remove parts of a recording that make the message harder to follow.

This may include:

  • Repeated ideas
  • Abandoned sentences
  • Filler words
  • Long pauses
  • Verbal stumbles
  • Unnecessary phrases

The tool can create a cleaner first pass, but it should be used carefully.

A completely polished conversation can sound unnatural. People pause, laugh, rethink an answer, and search for the right words.

Those moments can make a podcast feel honest.

The best outcome is not a conversation with every “um” removed. The best outcome is a conversation that is easy to follow while still sounding human.

For an emotional interview, keep pauses that allow the guest’s words to settle.

For an educational video, you may want a faster pace with tighter sentences.

Your editing request should explain the type of result you want.

Better Filler-Word Removal

Descript can identify words such as “um” and “uh” and let you remove them in groups or one at a time.

Filler-word removal now supports more languages, including Spanish, German, French, Portuguese, and Italian.

Descript also updated its word-level alignment to produce smoother results when transcript edits are applied to clear recordings.

This can help cuts sound less choppy.

Do not automatically remove every filler word.

Removing a filler word between two fast sentences may sound natural. Removing one in the middle of a thoughtful answer may create an abrupt cut.

A good workflow is to let Descript identify the filler words, remove the obvious distractions, and then listen to the remaining cuts.

Improved Transcription

Accurate transcription is important because the transcript controls many of Descript’s editing tools.

Descript added ElevenLabs Scribe v2 as a transcription option and later added Rev AI v3. It also provides a model picker so users can choose an available transcription engine.

Better transcription can mean:

  • Fewer spelling corrections
  • Better sentence structure
  • Improved speaker identification
  • More accurate searches
  • Cleaner captions
  • Faster editing

Names, companies, book titles, and technical terms may still need manual correction.

Correct important names before using the transcript to generate show notes, descriptions, or captions. If a guest’s name is wrong in the transcript, that mistake may appear in every piece of promotional content created from it.

Studio Sound Helps Rescue Weak Audio

Studio Sound is one of Descript’s most practical features.

It can reduce background noise and improve the clarity of spoken audio. This can help when someone records in a room with echo, uses a basic microphone, or has background noise in the recording.

It can be useful for:

  • Remote interviews
  • Webcam recordings
  • Home offices
  • Rooms with hard walls
  • Quiet background hum
  • Uneven voice quality

The desired outcome is a voice that sounds clearer and easier to understand.

More processing is not always better.

Applying Studio Sound too strongly can make a voice sound artificial. Start with a moderate setting and compare it with the original audio.

Pay close attention to laughter, breathing, emotional speech, and words spoken away from the microphone. These parts can react differently to heavy audio processing.

Studio Sound can improve a usable recording. It cannot fully replace proper microphone placement and a quiet recording space.

Regenerate Speech Can Fix Small Mistakes

Regenerate lets you replace or correct short spoken sections.

You may be able to fix:

  • A mispronounced word
  • A wrong date
  • A missing word
  • A short verbal mistake
  • A small change in phrasing

Descript says its newer Regenerate model is designed to better preserve natural pacing and intonation.

This can help replacement audio blend with the surrounding sentence.

Regenerate is most useful for small repairs. Replacing a full paragraph may be more noticeable and may change the natural rhythm of the speaker.

You should also have permission before creating or using another person’s AI-generated voice.

For a host fixing a mistake in their own introduction, this feature can save a complete re-recording.

Improve Lighting Inside Descript

Poor lighting is common in remote podcast interviews.

A guest may sit in front of a bright window. A room may be too dark. Two cameras may have completely different colors.

Descript has expanded its color controls and video effects.

The color-adjustment tools include:

  • Exposure
  • Contrast
  • Highlights
  • Shadows
  • Saturation
  • Temperature
  • Tint
  • White balance

There are also presets such as:

  • Neutral
  • Warm
  • Cool
  • Pop
  • Black and White

These tools can help you make footage from different cameras look more consistent.

Descript also introduced Portrait Lighting and Uplighting.

Portrait Lighting is designed to keep the subject brighter while darkening the background. Uplighting can brighten footage recorded in low light.

These effects use AI processing and may use AI credits, depending on the feature and plan.

The desired outcome is not a heavily edited appearance. You want the speaker’s face to be clear and the footage to look intentional.

New Visual Effects for Creative Videos

Descript added several effects to its Effects drawer.

Options include:

  • Vignette
  • Selective Color
  • Pixelate
  • Color Tone
  • VHS
  • Invert
  • Threshold
  • ASCII styling
  • Portrait Lighting
  • Uplighting

Some of these effects are useful for branding or storytelling. Pixelate could help lightly hide sensitive information. Vignette can bring attention toward the middle of the frame. VHS can create a retro look for a specific segment.

Use visual effects with a purpose.

A strong story, clean sound, readable captions, and a clear opening will usually do more for a video than several effects added without a reason.

Create Promotional Clips From a Full Episode

A full podcast episode contains more content than the finished episode itself.

One interview may include several quotes, stories, questions, and lessons that could work as separate videos.

Descript’s Create Clips tool can help find possible short-form moments in a longer recording.

You can then adjust:

  • Clip length
  • Opening hook
  • Layout
  • Framing
  • Captions
  • Background
  • Font
  • Brand colors
  • Aspect ratio

You could turn one interview into:

  • One full podcast episode
  • One YouTube video
  • Three short vertical videos
  • One episode trailer
  • One quote video
  • One audiogram
  • One LinkedIn clip
  • One email topic

The strongest clip is not always the loudest or most dramatic moment.

Look for a section that makes sense on its own. The viewer should understand the topic without watching the entire interview.

A good clip often begins with:

  • A surprising statement
  • A clear problem
  • A personal confession
  • A practical tip
  • A question the audience already has

Let the AI identify possible clips, then use your knowledge of the audience to choose the final ones.

Generate Captions and Subtitles

Captions help people follow videos without sound. They also make content easier to understand in noisy places and more accessible for people who are deaf or hard of hearing.

Descript can generate captions from the transcript and place them on the video.

You can adjust the caption style to match your content.

Recent updates added gradient fills for captions and other visual elements. Descript also supports word-level timestamps in subtitle exports, which can provide more precise timing when captions are moved into another editing program.

Keep captions easy to read.

Use:

  • A clear font
  • Strong contrast
  • Short lines
  • A readable size
  • Careful placement
  • Correct spelling

Review names, places, and industry terms before exporting.

Animated captions may help short videos hold attention, but captions should support the speaker instead of competing with them.

Generate YouTube Thumbnails Inside the Project

Descript now offers an option to generate YouTube thumbnails.

You can run it as an AI action or ask Underlord to create a thumbnail from your project.

This could reduce the need to move between several tools when preparing a video.

A generated thumbnail still needs a human review.

Check:

  • Is the text readable on a phone?
  • Does the image match the video?
  • Is the person’s face accurate?
  • Is the main idea clear?
  • Is there too much text?
  • Does it look like the rest of your brand?
  • Does it make a promise the video actually fulfills?

The goal is not just to make an attractive image. The thumbnail should give the right person a reason to click without misleading them.

Edit an Image by Drawing on It

Descript added AI image editing that lets you draw over a specific area and describe the change.

You may be able to:

  • Remove an object
  • Replace a background
  • Change part of the sky
  • Remove a logo
  • Clean up a distracting section
  • Replace an item in the frame

Descript then attempts to repaint only the selected region.

This can help when a thumbnail, title card, or piece of B-roll contains one distracting object.

Review AI-edited images closely. Look at faces, hands, text, logos, and edges around the changed area.

Use the Media Library Across Projects

Descript’s Media Library lets creators store and organize reusable assets at the Drive level.

You can upload:

  • Video clips
  • Audio files
  • Images
  • Music
  • Intros
  • Outros
  • Logos
  • Backgrounds
  • Sound effects

You can then access these files from multiple projects.

This can save time if every podcast episode uses the same intro music, logo, guest frame, or call-to-action video.

Files uploaded to the Media Library can be organized into folders and searched by filename. Descript also added automatic transcription for audio and video uploaded to the library.

Reusing a previously uploaded Media Library file in a project does not require another transcription of the same file, according to Descript’s changelog.

For a regular podcast, this can make the weekly workflow more organized.

Schedule Remote Recording Rooms

Descript Rooms is its remote recording system.

You can invite guests to record audio and video from different locations. Each participant’s recording can be captured separately, giving you more control during editing.

Descript now lets you schedule a Room in advance and add the recording link to Google Calendar or Outlook.

This removes one small but common problem. You do not have to create a new recording link minutes before the interview.

Guests can receive the correct link with the meeting invitation.

Descript has also worked on:

  • Recording recovery
  • Low-storage warnings
  • CPU-overload warnings
  • Live captions
  • Camera reconnection
  • Audio-dropout fixes
  • Upload reliability
  • Long-recording stability

These updates may not sound as exciting as generative AI, but recording reliability is more important when you have one chance to interview a guest.

Live Captions in Recording Rooms

Descript Rooms supports live captions during a session.

Participants can read an on-screen caption overlay while the conversation takes place.

This may help:

  • People who have trouble hearing
  • Guests in noisy environments
  • Participants following unfamiliar terms
  • Fast-moving group conversations
  • Hosts checking whether speech is being detected

Live captions are not the same as the final corrected transcript. They can still make the conversation easier to follow during recording.

Create Show Notes and Promotional Content

Once the episode is edited, Underlord can help produce first drafts of the content surrounding it.

You can ask for:

  • Episode titles
  • Show notes
  • Chapter titles
  • YouTube descriptions
  • Social posts
  • Short-video hooks
  • Email subject lines
  • Newsletter copy
  • Blog outlines
  • Quotes
  • Key lessons
  • Guest summaries

This can save time, but the draft should not be published without review.

AI-written promotion can sound generic if the request lacks context.

Tell Underlord:

  • Who the audience is
  • What problem the episode addresses
  • What the guest teaches
  • What emotional tone you want
  • Which story should lead
  • What action the listener should take
  • What phrases to avoid

You can also provide examples of your own writing.

The tool should help you develop a starting point. Your personal voice and knowledge of the episode should shape the final version.

Connect Descript With Other AI Tools

Descript has released an API in open beta and supports Model Context Protocol connections.

This can allow compatible AI assistants and outside workflows to interact with Descript.

Possible actions include:

  • Importing media
  • Searching projects
  • Targeting a composition
  • Editing with Underlord
  • Generating content
  • Checking task progress
  • Triggering publishing

This is a more advanced feature. Most creators will not need it to edit a weekly podcast.

It may become useful for someone building a repeatable content-production system.

A future workflow could take a new recording, create a project, complete a first-pass edit, prepare promotional assets, and organize the results for review.

Descript warns that the API is in beta, so its behavior and available endpoints may change.

A Practical Descript Workflow for Podcasters

Here is a simple workflow that keeps the creator in control.

Step 1:Import the recording

Upload your audio or video to a new project.

Make sure all participant tracks are present before you start editing.

Step 2:Correct the transcript

Correct the guest’s name, company, book title, website, and any important terms.

This improves captions and AI-generated promotional content later.

Step 3:Save an original version

Use version history or duplicate the composition before making large AI edits.

This gives you a way to return to the original structure.

Step 4:Ask Underlord for a first pass

Give it a detailed editing request.

Example:

Create a first-pass edit of this podcast interview. Remove technical setup, false starts, repeated questions, long pauses, and repeated answers. Keep the guest’s personal stories, practical advice, natural laughter, and emotional pauses. Do not change the meaning of any answer.

Step 5:Review the suggested changes

Read the transcript and check any section where the meaning may have changed.

Restore parts that feel too aggressive.

Step 6:Improve the audio

Apply Studio Sound carefully. Level the voices and check the transition between speakers.

Listen with headphones before exporting.

Step 7:Improve the video

Correct obvious lighting problems. Review camera layouts, captions, and jump cuts.

Use regenerated video only where the result looks natural.

Step 8:Watch the full episode

Do not skip this step.

Watch or listen from beginning to end. Check for missing words, abrupt cuts, audio problems, and repeated sections.

Step 9:Create promotional clips

Ask Descript for several options, then choose the strongest three.

Give each clip a clear opening and a single main idea.

Step 10:Generate draft promotional material

Create show notes, titles, descriptions, chapters, and social copy.

Edit everything into your own voice before publishing.

Step 11:Export and check the final files

Open the exported files and inspect the beginning, middle, and end.

Confirm the resolution, sound, captions, and total length.

Descript Prompts You Can Try

Podcast interview prompt

Edit this interview for clarity and pacing. Remove technical setup, false starts, accidental repetition, filler words that distract from the message, and pauses longer than two seconds. Keep emotional pauses, natural laughter, personal stories, and the guest’s original meaning. Flag any section where a cut may change the context.

YouTube editing prompt

Turn this recording into a focused YouTube video for adults over 40 who are learning AI. Keep the language simple. Start with the strongest problem or question. Remove repetition and unrelated sections. Aim for eight to ten minutes. Suggest places for captions, B-roll, and pattern changes.

Short-video prompt

Find five moments that can work as 30 to 60-second vertical videos. Each clip must make sense without the full episode. Prioritize surprising lessons, practical tips, honest stories, and clear statements. Give each clip a short hook and suggested title.

Show-notes prompt

Write podcast show notes in a friendly, natural voice. Start with the listener’s problem. Explain why this conversation matters. Include five key takeaways, a short guest introduction, and a simple call to action. Do not use exaggerated claims or generic marketing language.

Final quality-check prompt

Review this edited episode for abrupt cuts, repeated ideas, missing context, uneven pacing, long host questions, weak transitions, and sections that may confuse a new listener. Do not make changes yet. Give me a list of suggested fixes with timestamps.

What Descript Still Cannot Replace

Descript can help with many parts of production, but it cannot replace your judgment.

It does not fully understand:

  • Your relationship with the guest
  • The history behind a personal story
  • What your regular audience already knows
  • Which pause carries emotional weight
  • Whether a joke fits your show
  • Which part represents your values
  • What you promised the guest
  • The larger direction of your podcast

AI can remove words. It cannot always understand what those words mean to the people involved.

Use Descript to reduce repetitive work.

Keep control of the message.

Is Descript Worth Trying in 2026?

Descript may be a good fit for:

  • Podcasters editing interviews
  • YouTube creators
  • Coaches and consultants
  • Small business owners
  • Course creators
  • People making talking-head videos
  • Beginners who find traditional editing software confusing
  • Busy creators who want to repurpose long-form content

It may be less suitable for projects that require advanced color grading, complex animation, detailed sound design, or film-level timeline editing.

Some AI features also use credits. Usage limits depend on your plan, and unused credits may not roll over. Review the current plan details before paying for a subscription. Check Descript’s current pricing and plan limits.

Descript’s main value is not that it does every edit perfectly.

Its value is helping you get from a rough recording to a strong first draft faster.

That can make the difference between an episode sitting unfinished on your computer and reaching the people who need to hear it.

Final Thoughts

Descript is moving beyond basic transcript editing.

Its newer features can help creators smooth jump cuts, improve weak lighting, clean up audio, remove distracting filler words, create clips, generate thumbnails, schedule remote interviews, and prepare promotional content.

Underlord also gives creators a new way to work. Instead of learning where every tool is located, you can explain the outcome you want in plain language.

That does not mean pressing one button and publishing whatever the AI produces.

The best workflow combines AI speed with human review.

Let Descript complete the repetitive first pass. Then use your ears, eyes, and knowledge of the audience to shape the final story.

For podcasters and ordinary creators, that is the real benefit. You spend less time fighting with editing software and more time creating useful content.

Affiliate disclosure: The following button uses my Descript affiliate link. If you sign up or purchase a plan through this link, I may earn a commission. You will not pay an additional fee because of the referral.

Try Descript and start editing your next podcast or video.

This article was researched and updated on August 11, 2026. Features, pricing, plan limits, and AI-credit requirements may change. Review the official Descript changelog and Descript Help Center for the latest information.

Joe Foley
Written by

Joe Foley

Joe Foley is the creator of AI for Ordinary People. He helps beginners, parents, creators, and small business owners use AI in simple, practical ways. Joe has been podcasting since 2013 and creates plain-English guides, prompts, and workflows for people who want useful help without tech jargon. His goal is to make AI feel less confusing and more useful for real life, real work, and everyday decisions.