NEW PODCAST: Wood vs. Silicon Valley: A DIY Digital Bolex, OpenAI Buys the Portrait Mode Team – Focus Check ep135 →🎙️ WATCH/LISTEN Now
Watch/Listen NowA DIY Digital Bolex & OpenAI Buys the Portrait Mode Team🎙️NEW PODCAST
Education for Filmmakers
Language
The CineD Channels
Info
New to CineD?
You are logged in as
We will send you notifications in your browser, every time a new article is published in this category.
You can change which notifications you are subscribed to in your notification settings.
Adobe has moved the After Effects AI Assistant from a beta-forum reveal to a formal public beta, promoting it alongside its IBC 2026 Premiere announcements and demonstrating it live on the show floor. A mid-August update addressed the directory-access issue, localized the panel, added text-layer editing, and improved how expression references are handled when layers are renamed. Early testers have also begun reporting on speed and reliability.
At IBC 2026 we gave the feature our CineD Best-of-Show Award in the category AI Innovation, Adobe’s second win in that category after Premiere Color Mode at NAB 2026. What convinced us was where Adobe chose to start: expression writing and inherited-project cleanup are two of the least glamorous jobs in motion design, and Adobe aimed the Assistant at both before anything flashier, while publishing its list of limitations at greater length than its list of capabilities.
The Assistant lives in After Effects (Beta), build 7 or later of the current beta stream, and is reached through the icon in the header bar, via Window, then Assistant, or through Quick Apply in the Timeline panel. It is opt-in, and Adobe has asked testers whether it should sit in a default workspace in the future.
Adobe’s product blog post positions the Assistant as the After Effects counterpart to the AI Assistant already in the Premiere beta and frames it as a creative agent that “works across an entire project, not just the composition you have open.”
The stated scope is reorganizing projects with thousands of layers, repairing technical issues, writing expressions from plain-language descriptions, and running inspection tasks across compositions and footage on request. Adobe names project organization and expressions as the two areas where it has spent the most engineering time; spell-checking and translating text layers, generating multiple aspect-ratio versions of a comp, and Firefly image and video generation run from the same panel with what Adobe calls varying success rates.
A detail from Adobe’s help documentation matters for anyone weighing the panel on client work: the Assistant’s settings menu offers two permission modes, “Always ask permission,” which requires approval for each action, and “Approve permission automatically,” which lets it proceed without stopping.
Early testers raised questions about how much of the local file system the Assistant could inspect. In one case, a request involving offline frames caused the Assistant to scan the footage location on disk rather than only inspecting the composition, prompting concerns about file access and permissions. Adobe responded that the Assistant was intended to request permission for actions like this, and later marked the issue resolved, saying the Assistant no longer scans or enumerates files outside the project when resolving paths and stays scoped to the task at hand.
Adobe also says it does not train on user content, keeps chat content only temporarily for bug investigation, and captures a single frame screenshot only when its project-vision tool needs one. Users working under NDA are advised to use their own discretion.
Beyond the scoping fix, the update addressed tools dropping out mid-session, the Assistant failing to connect when the panel was open at startup, and panel loading reliability. The panel is now fully localized, so the earlier limitation of working best in English, where tool calls failed in other languages, is listed as resolved.
New capabilities in that build include text manipulation, so the Assistant can edit and drive text layers directly; an Open Comp action that lets it open the composition it is working in; and direct file generation and import instead of linking to help documentation. Adobe also lists a set of accuracy improvements on real work. Renaming a layer no longer breaks expressions that reference it, because the Assistant auto-fixes the references. The correct 3D renderer is chosen for parametric and mesh scenes. Shape-layer hierarchies are handled better, including selection awareness, nested groups, and driving the right content property, for example a Rectangle Path’s Size rather than the layer’s Scale. Checkbox-driven expressions now include the required .value, and Adobe reports fewer incorrect “that’s out of scope” responses.
Early feedback in Adobe’s beta thread points to speed and reliability as the main rough edges, but testers also see potential. One user reported that an image-sequence task took more than ten minutes of processing and back-and-forth before the final action, although once the Assistant understood the task, the action itself was instant.
Adobe’s own list of remaining limitations includes continued speed optimization, reference uploads currently limited to JPG and PNG, scripts that can be generated but not executed, and undo operations that are handled individually rather than grouping an entire request.
The Assistant is available now in After Effects (Beta), installed through the Beta apps tab of the Creative Cloud desktop app. The beta is free, including Firefly generations, subject to a daily limit and a separate weekly threshold that Adobe has not quantified. Adobe says it will share a pricing plan for the full release when it has one. The shipping application remains After Effects 26.5, the September 2026 release, which does not include the Assistant. For more information, please visit the Adobe After Effects website.
Would you run the Assistant on client projects after this fix round, and where has it saved or cost you time so far? Don’t hesitate to let us know in the comments below!
Δ
Stay current with regular CineD updates about news, reviews, how-to’s and more.
You can unsubscribe at any time via an unsubscribe link included in every newsletter. For further details, see our Privacy Policy
Want regular CineD updates about news, reviews, how-to’s and more?Sign up to our newsletter and we will give you just that.
You can unsubscribe at any time via an unsubscribe link included in every newsletter. The data provided and the newsletter opening statistics will be stored on a personal data basis until you unsubscribe. For further details, see our Privacy Policy
A camera was put in my hands at 16, and I’ve been taking photographs ever since. Before moving to Vienna, I worked as a photographer on film sets, telling stories as they unfolded behind the scenes. Since then, I’ve focused on immersive theater, both as a photographer and an actor. I am the photographer for the award-winning Nesterval ensemble, documenting the energy and atmosphere of their performances. I’m a teller of stories. Stories hold my interest. I’ll take on any project with a story to tell.