AMD Releases Lemonade 11.0 Local AI Server with Text-to-Speech and New Features (github.com)

AMD has unveiled the Lemonade 11.0 Local AI Server, featuring significant improvements such as a new text-to-speech (TTS) capability that incorporates OpenMOSS for voice-cloning and voice-design models. Additionally, the release introduces a 3D-generation modality that ports the Trellis.2 image-to-3D pipeline, enhancing user interaction with a dedicated GUI 3D panel and a new endpoint for generating 3D content. This server also supports ModelScope as a secondary remote model registry alongside Hugging Face, streamlining access to a broader range of AI models. The significance of this release lies in its comprehensive enhancements that cater to a wide array of AI developers. The auto-installation feature for the FastFlowLM NPU backend on Linux simplifies setup, while adjustments to the model registry prevent overwriting existing user models, thereby safeguarding user data. Technical aspects like the Docker runtime changes, support for AMD Instinct MI300X, and various security enhancements improve the operational reliability and usability of the platform. Overall, these updates position the Lemonade Local AI Server as a versatile tool in the evolving AI landscape, allowing developers to create more immersive applications.