ElevenLabs released Music v2.5, its latest AI music generation model, available through both app and API. The new version outperformed its predecessor in blind testing with nearly 48,000 comparison pairs.
ElevenLabs has rolled out Music v2.5 for users seeking AI-generated audio tracks. The model demonstrates measurable improvements, with listeners preferring it over the previous version in large-scale testing across nearly 48,000 comparison pairs.
The company trained Music v2.5 exclusively on licensed music, addressing concerns around data sourcing in generative AI development.
Access is available through two pricing models: a free tier for basic users and a pro tier for those requiring advanced features or higher usage limits. Both the web app and API integrate the new version, enabling developers and creators to incorporate Music v2.5 into their workflows.
This release positions ElevenLabs within the competitive AI music generation market, where multiple companies offer similar tools for creating original compositions. The emphasis on licensed training data reflects industry movement toward addressing copyright considerations in AI model development.
Satya Nadella announced a formal Code of Conduct for Microsoft's multimodal AI (MAI) models, emphasizing the company's commitment to deliberate development that prioritizes human control and alignment.
AllSpark has unveiled Iris-mini and Iris-pro, open-source search agents built on Qwen models that top benchmarks among open-weight competitors in their respective size classes.
The UK government is piloting three-week AI training bootcamps for 16- to 24-year-olds in Preston, with apprenticeship placements offered to participants upon completion.