Descript is a text-based video and audio editing platform that revolutionizes podcast and talking-head video production by letting users edit media by editing a transcript — delete a word in the text, and the corresponding audio/video segment is removed automatically. As of mid-2026, the platform's AI layer includes Studio Sound (AI noise removal and audio enhancement), Overdub (text-to-speech voice cloning), Eye Contact correction, automatic filler-word removal, caption generation, and the Underlord AI co-editor for automated editing suggestions. The workflow is genuinely transformative for speech-heavy content: a 60-minute podcast can be rough-cut in 20 minutes by editing the transcript rather than scrubbing a timeline. Pricing runs across five tiers: Free ($0, 1 hour transcription/month, watermarked exports, basic features), Hobbyist ($16/month billed annually or $24/month-to-month, 10 hours transcription, watermark-free 4K export, no Studio Sound or Underlord), Creator ($24/month annually or $35/month-to-month, 30 hours transcription, full Studio Sound, Eye Contact, Underlord access, 800 AI credits), Business ($50/month annually or $65/month-to-month, 40 hours transcription, priority support, advanced collaboration), and Enterprise (custom). Annual billing saves roughly 25–35%. Transcription overages cost approximately $1 per extra hour, billed in 6-second increments — meaning re-importing a file three times due to export issues charges three times. User feedback is sharply divided by use case. Podcasters and YouTubers call it "the best editor for talking heads" and report cutting editing time by roughly 50%. Studio Sound is genuinely praised as a miracle for noisy recordings. However, transcription accuracy sits at ~96% on clean audio and drops noticeably on accents, technical jargon, and overlapping speakers — misaligned transcripts can produce wrong cuts that clip words the user wanted to keep. Overdub voice cloning is widely described as "robotic on long passages" and unsuitable for replacing real narration. The platform is also unsuitable for cinematic editing, music videos, or VFX-heavy work — "the worst editor for anything cinematic" is a common verdict. For speech-heavy content, Descript is a genuine productivity multiplier; for visual storytelling, traditional NLEs remain essential.