Adobe Firefly is a web application and family of generative artificial intelligence models developed by Adobe for creative production. Its capabilities include text-to-image, text-to-video, image-to-video, speech generation, soundtrack generation, sound effects generation, and speech enhancement. Firefly is part of Adobe Creative Cloud and also powers features in other Creative Cloud applications, including Photoshop, Premiere Pro, and InDesign. It is available via subscription plans that use monthly credits based on the type of AI being used.
Firefly is developed using Adobe's Sensei platform, the company's machine learning and artificial intelligence framework. The models are trained with images from Creative Commons, Wikimedia, and Flickr Commons, as well as 300 million images and videos from Adobe Stock and the public domain. This reliance on training data for which Adobe owns the license or which is in the public domain led the company to describe the models' output as "commercially safe" and legally owned or licensed. However, it was later revealed that Firefly was also trained on images from Midjourney and other competitors, which raised questions about those claims.
History
Adobe Firefly was first announced in September 2022 at Adobe's MAX conference. It was initially released as a public beta in March 2023 and is currently available to all Adobe Creative Cloud subscribers. Firefly is built on top of Adobe Sensei, which has been used to power various features in Adobe's creative software, such as object selection in Photoshop and image auto-enhancement in Lightroom.
Firefly expanded its capabilities to Illustrator, Premiere Pro, and Express, particularly for generating photos, videos, and audio to enhance or alter specific parts of media. Google planned to use Firefly in Bard (now Gemini) as its AI image generator but ended up using its own Imagen model. Mattel, IBM, and Dentsu have also partnered with Adobe. The initial demo showed a capability for generating variations of photos.
In April 2025, Adobe released Firefly Image 4 and Firefly Image 4 Ultra Model in the Firefly web application. In October 2025, Adobe released Image Model 5, Generate Soundtrack, Generate Speech, and Firefly Video Editor, announced at the Adobe MAX conference.
Capabilities
Firefly supports multiple modalities, including text-to-speech, text-to-vector, image-to-video, video, image, audio, and music. These features allow users to generate and edit visual and audio content from textual descriptions, facilitating creative workflows across various media types.
Integrations
Adobe Firefly is hosted partially from the Nvidia Picasso generative AI library. Firefly and other Adobe products are integrated with Google's AI models, including Gemini, Veo, and Imagen. Firefly is also equipped with AI models from OpenAI and ElevenLabs. These integrations expand Firefly's functionality and allow users to leverage a broader range of AI capabilities.
Adobe Firefly is used by Mattel to design packaging for all toys and subsidiaries, including Barbie. IBM and Dentsu also have integrations with Adobe Firefly, demonstrating its adoption in enterprise and creative industries.
Training Data and Commercial Safety
Firefly's training data initially comprised images from Creative Commons, Wikimedia, Flickr Commons, Adobe Stock, and the public domain, totaling over 300 million images and videos. Adobe emphasized that this approach ensured the models' outputs were commercially safe, as the training data was either licensed or public domain. However, later revelations indicated that Firefly was also trained on images from competitors like Midjourney, which contradicted Adobe's earlier claims. This raised concerns about the legal and ethical implications of the training data.
Enterprise and Partnerships
Firefly for Enterprise was released on June 22, 2023, targeting business users with additional features and support. Partnerships with companies like Mattel, IBM, and Dentsu highlight Firefly's integration into large-scale design and marketing workflows. These collaborations demonstrate Firefly's utility in generating consistent visual content for branding and product design.
Reception and Impact
Firefly has been received as a significant step in making generative AI accessible to creative professionals, integrating seamlessly with existing Adobe tools. Its ability to generate and edit images, videos, and audio has been praised for enhancing productivity and creativity. However, the controversy over training data has sparked debates about transparency and copyright in AI development, influencing discussions within the broader AI community.
Future Developments
Adobe continues to iterate on Firefly, releasing new models and features regularly. The October 2025 release of Image Model 5 and the Firefly Video Editor indicates ongoing expansion into video and audio generation. As AI technology evolves, Firefly is likely to incorporate more advanced capabilities, potentially integrating with other emerging AI systems and platforms.
References
- Adobe Firefly - Wikipedia (source facts provided).