Description
Adobe Firefly today is more than Adobe's own model: a multi-model workspace for generative images, video, audio and vector graphics. The native Firefly models are still trained exclusively on licensed content, public domain material and Adobe Stock. Since March 2026, the same interface has also given access to more than 30 third-party models, including from Google, OpenAI, Runway and Kling. From training its own models, Adobe derives IP indemnification for enterprise use, a contractual safeguard against copyright claims. This safeguard applies to the genuine Firefly models; for outputs from third-party models inside Firefly, it is worth checking Adobe's current terms before commercial use. Business plans separately include a commercial license.
The strength of the tool lies in its integration into existing Adobe workflows: Generative Fill and Expand directly in Photoshop, text to video in Premiere, AI generated vectors directly editable in Illustrator, templates in Express. Anyone already working in Creative Cloud gets the AI features inside the programs they already use, without switching tools.
The genuine Firefly models are still considered more conservative and less experimental than standalone image generators such as Midjourney, but commercially safer. Anyone looking for stylistically more experimental results within the same workspace can now also use the third-party models integrated there. All generations run on a credit system, and the scope depends on the model and output chosen. For best results, English prompts are recommended; German prompts work, but with limitations.









