At Adobe Summit on April 20, 2026, Adobe expanded its NVIDIA partnership with two pieces aimed at brand and studio creators: Firefly Foundry, which fine-tunes Adobe's Firefly models on a brand or franchise's own IP, and a 3D digital twin solution built on NVIDIA Omniverse for scalable on-brand content generation. WPP is the launch agency partner, and CAA, WME, UTA, B5 Studios, and directors David Ayer and Jaume Collet-Serra are signed on to Firefly Foundry.

For the broader landscape, see our complete guide to AI image generation in 2026.

What Happened

NVIDIA announced the partnership expansion on April 20. The integration runs Adobe's creative and marketing agents on NVIDIA Nemotron models inside NVIDIA OpenShell, a secure runtime for policy-governed agent execution. Firefly Foundry sits on NVIDIA AI infrastructure for the actual model tuning. WPP is rolling out the joint stack across its global agency network as the proof point.

Alongside the creative side, Adobe also launched CX Enterprise Coworker, an agentic system that orchestrates downstream customer experience workflows. AWS, Anthropic, Google Cloud, Microsoft, and OpenAI are all listed as integration partners. CX Enterprise Coworker is generally available in the coming months.

Why It Matters

Firefly Foundry is the headline for working creators. Until now, a brand or studio that wanted custom AI image, video, and 3D models on its own IP had to either commission a fine-tune from a model lab or build the pipeline in-house. Foundry packages that as an Adobe service, with commercially safe outputs and indemnification baked in, running on NVIDIA infrastructure. For VFX shops, advertising agencies, and franchise IP holders, that removes a real chunk of the procurement and legal friction that has slowed enterprise AI adoption.

The talent agency presence (CAA, WME, UTA) signals where this is aimed: managing AI rights and AI-trained likenesses for clients. With YouTube expanding its likeness detection program last week and the WGA's four-year AI deal setting precedent, agencies need infrastructure that can train on a client's IP without leaking it. Foundry is the tooling layer for that.

Key Details

  • Firefly Foundry: Custom-tuned Firefly models on a brand or IP holder's proprietary content; runs on NVIDIA AI infrastructure; outputs commercially safe images, video, audio, 3D, and vector
  • 3D digital twins: Built on NVIDIA Omniverse and OpenUSD for scalable on-brand 3D content
  • CX Enterprise Coworker: Agentic AI for customer experience orchestration; integrations with NVIDIA Nemotron, AWS, Anthropic, Google Cloud, Microsoft, OpenAI
  • Runtime: Agents execute inside NVIDIA OpenShell, a policy-governed secure environment
  • Foundry partners: WPP (lead agency), CAA, WME, UTA, B5 Studios, Promise Advanced Imagination, Cantina Creative; directors David Ayer, Jaume Collet-Serra
  • Availability: CX Enterprise Coworker generally available in the coming months; Foundry partners onboarding now

What to Do Next

If you run a small studio or agency with original IP, watch how Foundry pricing lands. Adobe has not disclosed tiers yet, and the early lineup is enterprise-scale (talent agencies, hybrid studios, top directors). If pricing comes down to mid-market in 2026, Foundry could replace a chunk of what creative shops currently bolt together with ComfyUI custom training and external LoRA work.

If you're an individual creator, the more useful track is the Firefly AI Assistant that Adobe shipped on April 15. Foundry is enterprise pricing for now. The Assistant is included in standard Creative Cloud and gives you the multi-app orchestration without the model-tuning bill.