- Descript’s November 2026 update introduced Generative Layouts, which builds complete brand template packs from uploaded guidelines, and Underlord Skills, which lets users save custom editing sequences as reusable slash commands.
- Generative Layouts directly targets the pain point of consistent branding, though AI credit consumption adds a variable monthly cost heavy users should budget for.
Descript’s November 2026 update adds two features that address the parts of video production most likely to bleed time: brand consistency and repetitive editing sequences. Generative Layouts builds complete template packs from uploaded brand guidelines; Underlord Skills turns saved editing conversations into one-keystroke slash commands. Both sit inside the same editor users already run for audio and video.
The Text-First Foundation
Descript built its reputation on treating audio and video like a word document. Its AI transcribes a recording and you edit the text to cut the media, delete a sentence from the transcript and that footage disappears. The November update extends that logic into visual design and workflow automation. Rather than opening a separate design app to build title cards or lower thirds, creators feed Descript their brand guidelines and the AI generates a full layout pack. Rather than typing the same editing instructions into the AI chat repeatedly, they save those sequences as reusable commands.
Generative Layouts
Upload your logos, colour palettes and style preferences, and Descript’s AI produces a complete set of on-brand templates covering titles, lower thirds and overlays. The packs are fully editable, and the Underlord AI co-editor can refine them further from a plain-language prompt. For teams, every video stays visually consistent without anyone switching to Canva mid-project. For solo creators, polished assets take minutes rather than hours.
Underlord Skills
Underlord, Descript’s built-in AI assistant, can now retain what you teach it. Previously, applying a specific sequence of edits meant typing the same instructions each time. With Underlord Skills, you perform the edit once, save the conversation as a named skill, and invoke it later with a slash command. Repetitive tasks, stylistic choices and multi-step cleanup sequences all become one-keystroke operations, with no coding required. The result is closer to a personal macro library than a prompt-driven chatbot.
The Broader AI Layer
Both features sit on top of existing AI tooling: Studio Sound for background noise removal and voice clarity, automatic filler word removal, and Animate Text for caption motion. Descript has also improved its transcription model for speed and accuracy, which matters because every downstream feature depends on that foundation. The platform now runs Anthropic’s Claude Sonnet 5.5 by default through its Media Composition Platform and API, underpinning the more complex assistance across the editor.
Credit Costs and Real Gaps
Most AI features draw from a monthly credit allowance on paid plans. Generative Layouts, Underlord, Studio Sound and AI Speech all consume credits, and high-output users will likely exhaust their allocation before month’s end. Extra credits cost more. That variable pricing is worth modelling before committing, particularly for teams running consistent volume.
The text-based approach also has genuine limits. Frame-accurate colour grading and complex visual effects still favour traditional timeline editors. AI transcription accuracy is high but not perfect, so a review pass before publishing remains necessary. Those gaps are worth weighing against the efficiency gains the platform delivers on common editing tasks.
Where Descript Fits
Keeping audio editing, video editing, design and workflow automation inside a single tool cuts context-switching and the overhead of managing multiple subscriptions. Descript’s text-first interface gives it a distinct position: accessible to people who would never open a professional NLE, while producing output that reads as polished. The practical question the November update leaves open is whether the credit model scales affordably as output volumes grow, a constraint that matters more to high-volume teams than to occasional users.
