Adobe Firefly at NAB 2026 – From Location Scout to Premiere Timeline with AI

by Alexandra ThompsonApril 25^th, 2026

Subscribe to News - Software

At NAB 2026, we caught up with Adobe at their booth to look at what is new on the Firefly platform, the company’s all-in-one home for generative AI. While Firefly is increasingly woven into desktop apps like Photoshop and Premiere Pro, the firefly.adobe.com web platform is where Adobe is consolidating image, video, and audio models alongside its growing partner ecosystem. The clear standout in our walkthrough was Firefly Boards: an infinite canvas that promises to take filmmakers from rough concept to finished cut without ever leaving the browser, with direct handoff to Premiere when it is time to finish.

We have covered Adobe’s Firefly platform extensively, from its original 2023 announcement through the public beta of the Firefly Video Model, the arrival of partner models like Aleph, Topaz Astra, and FLUX.2, the Firefly Image Model 5 announced at Adobe MAX 2025, and most recently the broader NAB 2026 update bundle that introduced Color Mode in Premiere and Frame.io Drive. What we had not yet seen in detail was how the web-based Firefly platform consolidates these capabilities in one place, especially around Firefly Boards and the credits system. Anton Knoblach, Adobe’s product evangelist for Firefly, who worked as a filmmaker and cinematographer before joining the company, walked us through it on day three of the show.

Firefly Boards: an infinite canvas for visual thinking

Boards is, in Anton’s view, the most useful Firefly feature for anyone working in video. The concept is simple: an effectively unlimited canvas where you can generate, organize, edit, and connect images and clips in the same workspace. The pitch is partly about overcoming what he called the “terrifying” feeling of a blank canvas, and partly about closing the gap between a filmmaker’s mental image of a scene and what their crew or clients actually understand from a verbal description. The argument is that visualizing ideas early lets you redo images instead of redoing shoots.

We saw two core workflows. The first was a documentary or commercial scenario set in the Alps. From a single short text prompt, Anton generated location-scouting frames, including a drone shot of a village, a wide shot of a cabin, and alternative times of day, and then chained those into rough animatics through video generation, all on the same board. Generated assets can be loaded back into the prompt bar to make targeted edits, for example shifting time of day from morning to dusk without leaving the canvas. Real-world references, including imported photography and existing video, can sit alongside generated imagery, so a still from an actual location scout can be reworked to preview sunset light without waiting for the sun to set.

Filling production gaps with AI shots

The second workflow was more pragmatic, and arguably the more interesting one for working professionals. Anton showed a frame from a real production shot inside an LED volume, where one drone B-roll was missing because the entire shoot had taken place in a studio. Starting with a still frame from the volume shoot, he used image models first to extend the background into a clean, non-blurred plate, then transformed the perspective into an aerial view, before finally feeding that into a video model to generate a usable drone shot and a fly-through variation based on the same starting frame. Topaz upscaling, integrated directly into Boards, then takes the output up to 4K for delivery.

This is also where Adobe’s positioning gets interesting. Rather than pushing Firefly as a way to “make” entire films from prompts, the demo deliberately framed it as a tool for filling gaps in real productions, an approach Anton said Adobe has formalized in what he described as an ethical AI filmmaking course. The argument is that generation is most defensible when it is anchored to actual footage and used to bridge what could not be captured on set, not as a substitute for the shoot itself.

Models, partners, and commercial safety

Firefly is both a product and a family of models, and Adobe’s own models cover image, video, audio, and speech. According to Adobe, those models are commercially safe, trained only on Adobe Stock, other licensed content, and public domain material, with no customer content used in training. We have covered this approach in detail before, including in our interview with Adobe’s Rajan Vashisht about responsible AI.

In response to user demand, Adobe has also opened Firefly to a curated selection of third-party models, so creators can choose between Adobe’s commercial-safety guarantee and other models based on the project. The lineup now includes Veo 3.1, Nano Banana 2, Runway Gen-4.5, Luma AI’s Ray3, FLUX.2 from Black Forest Labs, and the recently added Kling 3.0 and Kling 3.0 Omni, with more than thirty models in the platform overall.

A practical example from the demo: a Veo 3.1 generation at 720p, 8 seconds, with audio costs 400 credits. The interface displays the cost before you commit to the generation, and the credit balance resets monthly, much like a mobile data plan. Native resolutions vary by model, with options including 540p, 720p, and 1080p, and the integrated Topaz upscaler can take results up to 4K.

Content Credentials and the AI ethics question

We pressed Anton on the recurring ethical concerns around generative video and image tools. His response framed Adobe’s position around two pillars: commercial safety in its own models, and transparency through Content Credentials.

Every asset generated in Firefly carries Content Credentials, regardless of whether it was created with an Adobe model or a third-party one. Through the Content Authenticity Initiative, the nonprofit Adobe co-founded, the credentials are designed to identify both that an image is AI-generated and which model produced it, and according to Anton, the verified tool can recover that provenance even from a screenshot. He also offered a more philosophical view, arguing that real creativity remains inherently human: average prompts will yield average output, and the bar for what counts as “great” is rising as a result.

Whether that reassurance is enough remains a personal call for each filmmaker, and at CineD we will continue to cover both sides of the debate so you can make your own decision.

Pitching, reviewing, and direct handoff to Premiere

One feature we were not expecting: Boards doubles as a pitch deck. Instead of exporting media into PowerPoint or Keynote, you can lay out a presentation directly on the canvas, share it with a client, and let them leave Frame.io-style comments on individual frames or rearrange items if they have edit access. From the same board, two clicks send selected media to Premiere Pro on the desktop without manual download and re-import, closing the loop between ideation and the actual NLE timeline.

Plans, credits, and Firefly-only access

Firefly is bundled into Adobe’s existing plans at different credit tiers. The most familiar entry point is Creative Cloud Pro, formerly known as Creative Cloud All Apps, which includes 4,000 credits per month. Add-on credit packs can be subscribed to and cancelled monthly when a project demands more. For producers, clients, or team members who do not need the desktop applications, Adobe also offers Firefly-only plans, with full pricing and per-model credit costs published on Adobe’s Firefly help pages.

Firefly Boards reframes generative AI in pre- and post-production as a place to think visually with collaborators rather than a black-box prompt machine, and the Premiere integration plus Content Credentials are a real attempt to address the most common professional concerns. Would you use it? Let us know in the comments below!