Descript is a video and podcast editor built around a transcript. Instead of cutting only on a waveform or timeline, you can remove words from the text and have the corresponding audio and video change with it. Its current AI assistant, Underlord, adds guided editing actions, layouts, clip creation, cleanup, and generation tools.
Text-based editing
Import or record media, let Descript transcribe it, then edit the transcript. Correct speaker names and technical terms before making large cuts. Removing a sentence can create an unnatural visual or audio jump, so preview transitions and use B-roll or room tone when needed.
Underlord
Underlord is Descript’s AI co-editor. It can help remove filler words and retakes, create clips, adjust gaze, improve sound, propose layouts, and apply workflows through templates. Use it on a duplicate composition and review each action rather than accepting a complete automated pass.
Studio Sound
Studio Sound reduces background noise and echo while enhancing speech. Apply it gradually: aggressive processing can make voices metallic, erase breath, or amplify artifacts. Compare with the original on headphones and speakers.
Scenes, layouts, and social clips
Scenes work like slide-based sections for B-roll, titles, multicam, and screen content. Create Clips can find short segments for social media, but a human should confirm that the excerpt keeps the speaker’s intended meaning and includes necessary context.
Screen recording and podcasts
Descript can record screen, camera, and microphone for tutorials or demos, and its transcript workflow suits interview podcasts. Verify notification privacy before screen recording and remove passwords, personal data, or private messages from the capture.
AI Speech and voice cloning
AI Speech can repair or generate spoken material and may support custom voice clones on eligible plans. Clone only your own voice or a speaker who has given explicit permission. Keep approval records and disclose synthetic speech if listeners could believe it was naturally recorded.
Plans and usage
Descript currently offers Free, Hobbyist, Creator, and higher options with different media hours, AI credits, export resolution, stock access, and Underlord capabilities. Official monthly and annual prices differ, so verify billing cadence and limits in checkout.
Recommended workflow
- Transcribe and correct the source.
- Duplicate the composition.
- Make structural edits in the transcript.
- Apply Studio Sound conservatively.
- Add scenes, B-roll, captions, and music.
- Review AI edits and export a test.
Limitations
Text editing is fast, but it cannot solve every visual continuity problem. AI credits, transcription errors, complex multicam work, advanced color, and detailed sound mixing may require other tools or manual editing.
Verdict
Descript is strongest for talking-head video, podcasts, tutorials, interviews, and social repurposing. Its transcript-first approach reduces routine work, while final pacing, accuracy, consent, and polish remain editorial decisions.