LOADING...
Summarize
Adobe's new AI tool instantly changes emotion in voiceovers
Adobe's Corrective AI is an improvement over Firefly's generative speech feature

Adobe's new AI tool instantly changes emotion in voiceovers

Oct 30, 2025
11:42 am

What's the story

Ahead of its MAX Sneaks event, Adobe has previewed a groundbreaking new AI tool called 'Corrective AI.' The innovative technology can change the tone and style of a voiceover in seconds. During a recent demonstration, WIRED witnessed the transformation of a lackluster voiceover into one that was confident and even whispered at times, just by selecting different preset emotions.

Evolution

Corrective AI builds on Firefly's generative speech feature

The Corrective AI tool is a step forward from the generative speech feature Adobe introduced for Firefly at this year's MAX conference. The earlier feature allowed users to use different preset voices and add emotion tags to change inflection. However, Corrective AI takes this capability further by allowing users to refine an existing voice performance instead of using a completely AI-generated one.

Audio separation

Project Clean Take can separate audio tracks

Along with Corrective AI, Adobe also showcased another Sneaks prototype called Project Clean Take. The tool uses AI to separate different parts of a single audio track into multiple ones. Although it is limited to five tracks, it can accurately separate voices, ambient noise, sound effects and more. This could be particularly useful for creators shooting in public spaces where licensed music is playing in the background.

Sound design

AI model adds sound effects to videos automatically

Adobe also demonstrated how its AI model can automatically analyze and add sound effects to a video. The tool breaks down the video into scenes, applies emotional tags and descriptions, and then generates relevant sound effects. While the demo wasn't perfect in every case, as per Wired, Adobe used a conversational interface similar to ChatGPT for manual adjustments instead of editing each scene individually.