Adobe Firefly announces new AI tools for video, audio & images

Adobe Firefly has introduced a suite of new artificial intelligence features, models, and partner integrations designed to streamline creative workflows across audio, video and imaging production.

The Firefly studio now offers creators tools for generating fully licensed audio tracks, voiceover production, and a timeline-based AI video editor, all accessible within a single platform. These releases expand Firefly's appeal to content professionals seeking to manage multimedia creation efficiently without leaving the Firefly environment.

Audio and video enhancements

The new Generate Soundtrack feature allows users to create AI-generated, studio-quality instrumental music that is fully licensed for commercial use. This tool produces multiple track variations in seconds and synchronises the audio precisely with video footage.

Accompanying this is the Generate Speech tool, a text-to-speech service supporting a range of lifelike voices, with multilingual capabilities and fine-tuning for emotion, pacing, and emphasis. The Speech tool uses both Firefly's proprietary Speech Model and voices from ElevenLabs.

Adobe has also added a web-based, multitrack video editor to Firefly. The editor supports generating, organising, trimming, and sequencing clips-all within the browser. It enables creators to add titles, soundtracks, and voiceovers, and to upload their own media for combining with generative content. The editor includes built-in style presets and a library of customisable title templates.

New image generation models

One of the key advancements is the introduction of Firefly Image Model 5, now available in public beta. The model produces photorealistic images in native 4MP resolution and supports prompt-based editing in everyday language. Image Model 5 achieves notable improvements in portrait anatomy, lighting, and multi-layered composition, and will power forthcoming features such as Layered Image Editing for in-depth, context-aware compositing.

Accompanying Image Model 5 is the 'Prompt to Edit' tool, which enables creators to specify changes to images using natural language-similar to text-to-image creation.

Partner and custom model ecosystem

The Firefly platform continues to add industry partner models. Newly integrated are ElevenLabs Multilingual v2 for voiceover generation and Topaz Bloom for image upscaling. These join various models from Black Forest Labs, Google, Luma AI, OpenAI, Runway, Topaz Labs, and others, giving creators broader options for content generation and editing within Firefly.

Adobe is also expanding early access to Firefly Custom Models, allowing users to train AI with their own image sets-provided they hold appropriate rights. Custom Models let creators build tools to generate assets that reflect their unique visual style. These custom models remain private by default and require the original creator's authorisation for use by others.

Ideation, production and workflow

Firefly Boards, the platform's collaborative ideation surface, now includes features such as Rotate Object to turn 2D images into 3D and new export options for streamlined collaboration. Additional recent tools include single-click image style presets and instant generative text editing within visuals.

Firefly Creative Production, now available in private beta, is aimed at professionals managing high-volume output. It provides batch editing capabilities for automatically replacing backgrounds, applying colour grading, and cropping images at scale, via a no-code interface.

Conversational AI agent

Adobe also previewed Project Moonlight, a conversational AI assistant integrated into Firefly. Project Moonlight can interact with Adobe's creative apps and creators' social channels to recommend content ideas and facilitate brainstorming sessions. The agent enables users to describe their objectives and desired outcomes in natural language, receiving personalised editing advice and suggestions while retaining creative control.

"Adobe Firefly supports your entire end-to-end creative process, from when you first explore ideas to when you bring them to the world with AI-powered video, audio, imaging and design," said Ely Greenfield, Chief Technology Officer and Senior Vice President, Adobe. "We built Firefly to be a single place where you can work with the best AI models in the best creative tools, and it's been amazing to see how creators are using it to bring their ideas from concept to reality."

Access and availability

Prompt to Edit with Firefly Image Model 5 and partner models is now generally available. Generate Soundtrack and Generate Speech are in public beta, and the Firefly video editor and Custom Models are currently rolling out in private beta. Firefly Creative Production is accessible to users with premium generative AI features, also in private beta. Project Moonlight is expected to be available in the coming months.

Adobe stated that through 1 December, Creative Cloud Pro and Firefly plan subscribers can access unlimited image and video generations with all supported models. Photoshop AI Assistant is also being introduced through a private beta for web users.

Adobe reaffirmed its approach to artificial intelligence, stating that it positions AI as a tool to support, not replace, human creativity, and that it remains committed to developing generative AI products in ways that respect creators' rights.