Release round-up—June 11th, 2026

Last month, the entire team got into a room, opened up your top-voted Canny requests, and built whatever we could in 48 hours. By the time we shut it down, and we'd donated . Many of the features below came directly out of the Telethon — they're called out as such in each section. For the full event recap, see Inside Descript's First Customer-Obsessed Telethon Hackathon.

Many of these updates require further design and testing before we can release them into the wild, so you'll be seeing these deploy throughout the summer.

You can now add inline tone directions to your AI speech scripts when using ElevenLabs V3 speakers. A Tone Tag is a small chip that appears directly in your transcript at the insertion point — it tells ElevenLabs v3 how to deliver the surrounding speech. Think whispers, laughs, pauses, excitement, or anything you want to describe in plain language.

To add one, place your cursor in the transcript and click in the toolbar. A preset picker appears with common options (happy, sad, whispers, laughs, pause, curious, and more). You can also type a custom direction — ElevenLabs treats it as a prompting instruction, so shorter and more specific tends to work better than long descriptions.

When you add or edit a tone tag, the surrounding paragraph regenerates automatically with the new tone applied. Paragraph breaks are the natural way to reset tone — the model determines how far the effect carries from each insertion point, so what you're really controlling is placement.

Tone Tags are available for . On any other speaker, inline notes remain regular in-line notes.

We've redesigned the Effects panel and added ten new effects to the lineup. The new Effects drawer lives in the Properties panel. Hover over any effect to preview it on the current frame, then click to apply.

The new effects:

*These items use AI processing and therefore use AI credits.

You can also access effects from the Properties panel on the sidebar.

See Apply and adjust effects and Visual and audio effects overview for full details on settings and parameters.

Shapes, text, scene backgrounds, and playback progress bars now support gradient fills. Pick from 21 presets or open Advanced controls to customize colors, stops, and direction. Underlord handles basic gradient instructions too — something like "apply a blue gradient to the Scene 1 background" works.

Here are some updates to Underlord since our last post.

Anthropic's newest model is available alongside the rest of the Underlord model lineup. Useful for complex, multi-step editing workflows where you want maximum reasoning quality.

Underlord's chat panel now renders markdown tables inline — useful for outlines, storyboards, and scene plans. Ask Underlord for structured content and it'll come back as a table you can scan.

Underlord now automatically detects and avoids unwanted letterboxing and crop bars in generated edits. Runs in the background — you don't need to do anything to enable it.

## Concurrent media generation

You can now run multiple AI media generation jobs at the same time without waiting for each to finish. Useful when you're iterating on b-roll, trying different prompts, or producing media for multiple scenes in parallel.

The Descript MCP connector is now live in two major AI assistant marketplaces. Connect your AI assistant to Descript and edit your projects, generate media, search Drive, and trigger publishes through natural language.

The Descript platform CLI is also updated. New commands include (export project transcripts from the command line), (browse Drive folders), and (list projects in your Drive). Update with .

Docs: http://docs.descriptapi.com

Viewers now enter Rooms sessions pre-muted — mic and camera stay off until they choose to publish. Business plan drives also get 10 Control Room producer slots, up from 3, now matching Enterprise.

Rooms hosts using Edge can now record participants in WAV, matching the lossless audio quality that's been available in other browsers. A direct sales ask from NPR, Shopify, McKinsey, and Microsoft.

When your microphone switches automatically mid-session (e.g. because a device reconnects), you'll now see a toast notification confirming what happened.

A new opt-in setting keeps your AI credits and minutes usage visible at all times. Off by default — open Settings and check the box to enable. Not available for enterprise drives; free drives already show usage by default.

An alternative color theme that minimizes red in the UI. The logo color in this theme is neutral (not red or lilac), per brand team feedback. Available in Settings.

We've made two upgrades to the "SuperToast" (the little pop up that lets you know something is running in the background, like file uploads etc.)

You can now cancel an upload at any point after adding files — while it's queued, processing, or mid-upload. And now, when you upload files, the SuperToast now shows individual progress indicators for each file — queued, uploading, transcribing, or failed. You can tell what's actually happening at a glance.

And there's even more:

Speaker detection now runs proactively on Drive media files in the background. Importing from Drive into a project means is available immediately — no waiting for detection to finish.