Back to stories
Tools

Adobe Launches Firefly Custom Models in Public Beta, Letting Creators Train AI on Their Own Art Style

Michael Ouroumis2 min read
Adobe Launches Firefly Custom Models in Public Beta, Letting Creators Train AI on Their Own Art Style

Custom AI That Looks Like You

Adobe has released Firefly Custom Models in public beta, giving creators and brands the ability to train personalized AI image generators on their own visual assets. The feature, announced on March 19, represents one of the most significant moves yet to bridge the gap between generative AI and brand-consistent creative production.

Users can feed the system 10 to 30 of their own images to create a custom model optimized for character design, illustration, or photographic styles. Adobe estimates training takes anywhere from 30 minutes to a couple of hours, depending on the model's complexity.

Preserving Visual Identity at Scale

The core value proposition is consistency. Custom models are designed to preserve details like stroke weight, color palettes, lighting, and character features across generations. This means a brand can explore new creative directions — campaign variants, product mockups, social media assets — without losing the visual identity that took years to build.

Crucially, custom models are private by default. For agencies and enterprises working with proprietary brand assets or unreleased product designs, this privacy guarantee is essential.

A Multi-Model Creative Platform

Alongside the custom models launch, Adobe has expanded Firefly into a multi-model platform offering access to more than 30 AI models from across the industry. These include Google's Nano Banana 2 and Veo 3.1, Runway's Gen-4.5, Adobe's own Firefly Image Model 5, and Kling's 2.5 Turbo.

Adobe also announced a strategic partnership with Nvidia to deliver next-generation Firefly models and accelerate creative, marketing, and agentic workflows.

Why This Matters

Generative AI image tools have been criticized for producing output that is visually impressive but generically "AI-looking" — disconnected from any particular brand or artistic voice. Custom models address this directly by putting the creator's own work at the center of the training process.

For enterprise teams spending heavily on content production, the ability to generate on-brand visuals at scale without sacrificing consistency could dramatically reduce turnaround times and production costs. For independent artists, it offers a way to leverage AI as an extension of their existing style rather than a replacement for it.

The public beta is available now through Adobe's Firefly platform, with broader enterprise features expected in the coming months.

Learn AI for Free — FreeAcademy.ai

Take "Prompt Engineering Practice" — a free course with certificate to master the skills behind this story.

More in Tools

Microsoft Backtracks After VS Code 1.118 Stamps Copilot as Git Co-Author by Default
Tools

Microsoft Backtracks After VS Code 1.118 Stamps Copilot as Git Co-Author by Default

VS Code 1.118 quietly flipped a setting that adds 'Co-Authored-by: Copilot' to Git commits — even on machines with AI features disabled. After developer backlash on GitHub and Hacker News, Microsoft says it will revert the default in version 1.119.

2 days ago2 min read
Cloudflare and Stripe Let AI Agents Buy Domains and Ship Code Without Humans
Tools

Cloudflare and Stripe Let AI Agents Buy Domains and Ship Code Without Humans

A new open protocol from Cloudflare and Stripe lets AI coding agents create accounts, buy domains, take API tokens, and push apps to production with no human in the loop.

2 days ago3 min read
Microsoft's $99 'Frontier Suite' Goes Live: 365 E7 and Agent 365 Hit General Availability
Tools

Microsoft's $99 'Frontier Suite' Goes Live: 365 E7 and Agent 365 Hit General Availability

Microsoft 365 E7 and Agent 365 became generally available on May 1, 2026, bundling Copilot, E5, Entra Suite, and an agent control plane at $99 per user per month for enterprise AI deployment.

3 days ago3 min read