Home / Video & film production / Descript
DDescript, Inc., San Francisco (US)
Descript
Video and podcast editing through the transcript: deleting text cuts the footage, with the AI co-editor Underlord, Studio Sound, voice cloning and translation.
Shared spot12of 20 professional tools with a GDPR profileOur summary of what the provider publishes itself about your data: whether the data is held in the EU (2 points), whether there is a certification or GDPR guarantee (1), and whether the provider is based in the Netherlands or the EU (1). 3 or 4 points is Strong, 2 Good, 1 Limited, 0 Unknown. How we measure in video & film production
GDPR: Unknown
- Provider based outside the EU
PriceFree $0Free $0; Hobbyist $24/month ($16/month billed yearly); Creator $35/month ($24/month billed yearly); Business $65/month ($50/month billed yearly), per person; Enterprise on request
Free trialYes
BasedUSOutside the EU
ModelNot publicAmerican company (not EU), Californian law. Transcription via Rev and a self-hosted Whisper model, avatars via Hedra, custom voices via Google Cloud; the language model Underlord uses is not stated. Descript may use input and output to train its own models unless you opt out; third parties are contractually not allowed to train on it. Storage in the US (Amazon S3 or Google Cloud). SOC 2 Type II; data processing agreement for customers and transfer mechanism not found.
Last checked: 9 November 2026 · How we check
Is this your tool? Fill in missing detailsWhat it does
Editing via the transcript
Underlord: AI co-editor
Studio Sound
Removing filler words
Eye Contact and Green Screen
AI voices and voice cloning
Translating and dubbing in 30+ languages
Comparable in video & film production
AmberScriptfrom €10 per hour (pay-as-you-go)…
Trint€69 per user/month (Pro), €59 per…
Limecrafton requestAPAdobe PremierePremiere €26.43/month (annual plan…
ElevenLabsfree plan
ColossyanStarter freeUnderlying model
GDPR check for video & film production US company (not EU), Californian law. Transcription via Rev and a self-hosted Whisper model, avatars via Hedra, custom voices via Google Cloud; which language model Underlord uses is not stated. Descript may use input and output to train its own models unless you opt out; third parties may not train on it under contract. Storage in the US (Amazon S3 or Google Cloud). SOC 2 Type II; data processing agreement for customers and transfer mechanism not found. Many tools build on a large model from, for example, OpenAI or Anthropic. If in doubt, ask for the data processing agreement: it states where your data goes.
Questions about Descript
- Which AI model does Descript use?
- The maker of Descript does not publicly disclose which AI model sits underneath. US company (not EU), Californian law. Transcription via Rev and a self-hosted Whisper model, avatars via Hedra, custom voices via Google Cloud; which language model Underlord uses is not stated. Descript may use input and output to train its own models unless you opt out; third parties may not train on it under contract. Storage in the US (Amazon S3 or Google Cloud). SOC 2 Type II; data processing agreement for customers and transfer mechanism not found.
- How much does Descript cost?
- Free $0; Hobbyist $24/mo ($16/mo billed yearly); Creator $35/mo ($24/mo billed yearly); Business $65/mo ($50/mo billed yearly), per person; Enterprise on request. Check the current price on the provider's website; prices often change.
- Who is Descript for?
- Descript is aimed at video & film production, marketing, communication & media. It fits work such as editing and post-production: cutting via the transcript, removing silences and filler words, reframing footage for social media and cleaning up audio, subtitles, translation and dubbing: automatic transcription, subtitles in dozens of languages and AI dubbing with voice cloning or lip sync, generating and enhancing imagery: b-roll and shots from text or images, upscaling to 4K, removing noise and motion blur.
- What should I watch out for with AI in video & film production?
- Raw footage almost always contains personal data: the faces and voices of people on screen, and with voice cloning or face editing you are working with biometric data. Arrange consent from the people depicted and, for voice cloning or lip-sync, from the actor or speaker. Check per tool what happens to uploads. CapCut may use your input and output for training under its standard terms (not on Enterprise), VEED does so on the free plan, and Descript does so unless you opt out. Runway (Bytedance, Singapore) stores data in Malaysia, Singapore and the US and receives a worldwide, perpetual licence to your content. Adobe, Kapwing, Flawless, Moments Lab and Dubformer state that they do not train on your content, Happy Scribe only with your permission. For unreleased client material, choose a tool with a data processing agreement and, where possible, EU storage or local processing.
Every Friday: what changed in AI for video & film production
New models, prices, oversight and tools, with sources and what it means for you. One email a week, unsubscribe with one click.