Stable Audio 2.5 Now Available on fal
We’re excited to announce the release of Stable Audio 2.5, developed by Stability AI and now available through fal. the latest audio model from Stability and the first designed with enterprise-grade use cases in mind.
Stable Audio 2.5 delivers a leap forward in both quality and creative control, meeting the growing demand for dynamic audio compositions that adapt seamlessly to a brand’s unique identity.
Research from Ipsos shows that custom audio makes a brand eight times more memorable, yet only 6% of creatives leverage a sound identity today. For enterprises looking to use sound more strategically, this gap represents a massive opportunity. To truly extend brand presence across touchpoints, companies need audio that is high-quality, commercially ready, and flexible enough to scale.
With Stable Audio 2.5’s enterprise-grade features, creative teams now have access to advanced, customizable audio generation tools, empowering every campaign, product, or production to carry the right sound.
Check out:
What is Stable Audio 2.5?
Stable Audio 2.5 is the newest generation of Stability AI’s audio model, designed specifically for enterprise-grade creative workflows. It combines cutting-edge research, faster performance, and greater creative control to help brands and creators generate commercial-quality audio that adapts seamlessly to different contexts.
Trained on a fully licensed dataset and built with commercial safety in mind, Stable Audio 2.5 enables enterprises to move beyond generic sound and develop custom, brand-led audio identities across campaigns, platforms, and experiences.
Key Features: Faster generation, smarter composition, enhanced workflows
- 3-minute tracks in seconds: Using the Adversarial Relativistic-Contrastive (ARC) method pioneered by the Stable Audio research team, the model can generate tracks up to three minutes long in less than two seconds on H100 GPUs.
- Dynamic, structured compositions: Outputs now feature richer musical development, with multi-part sections such as intros, transitions, and outros. The model also responds more precisely to descriptive prompts, whether it’s a mood like “uplifting” or musical language like “lush synthesizers.”
- Audio inpainting for control: Beyond text-to-audio and audio-to-audio, Stable Audio 2.5 adds support for audio inpainting. Users can upload audio, select a starting point, and let the model generate the continuation—ideal for tailoring edits within an existing track.
Custom, brand-led audio with creative control
Audio shapes brand engagement by up to 86%, yet most enterprises lack a consistent sonic identity. Stable Audio 2.5 gives teams the tools to design intentional, on-brand soundscapes that scale across touchpoints, from ads and in-store experiences to product sounds like card chimes or automotive audio.
Try Stable Audio 2.5 on fal.ai
You can experience the power of Stable Audio 2.5 directly on fal. Generate professional-grade audio in seconds, explore dynamic compositions, and experiment with advanced workflows like audio inpainting, all from an easy-to-use platform built for speed and scale.