AMD releases Lemonade 11.0, a major update to its open-source local AI server, adding text-to-speech with voice cloning and voice design via the OpenMOSS backend — enabling entirely offline, privacy-preserving speech synthesis on consumer AMD hardware.
On July 15, 2026, AMD released Lemonade 11.0, the latest version of its open-source local AI server, with text-to-speech as the headline feature. The update integrates an OpenMOSS-powered TTS backend that supports both voice cloning — creating synthetic voices that mimic a specific speaker from a short recording — and voice design, which generates custom synthetic voices from scratch. A dedicated TTS panel built directly into the Lemonade desktop GUI makes both features accessible without command-line configuration, lowering the barrier for users who are not comfortable with terminal-based tools.
Lemonade is AMD's community-maintained, fully open-source alternative to cloud AI services, designed to turn a consumer PC into a private AI server. Version 11.0 runs on AMD Ryzen CPUs, Radeon GPUs, and Ryzen AI NPUs on both Windows and Linux, with compatibility also extended to NVIDIA GPUs, Apple Silicon, and standard CPUs depending on the model and backend. The privacy advantage is central to the pitch: because all voice cloning and speech generation happens locally on the user's own hardware, no audio data ever leaves the device — a sharp contrast to cloud-based TTS APIs from ElevenLabs, OpenAI, and Google, which require uploading text and reference audio to remote servers for processing.
The release arrives at a moment when local, on-device AI is gaining momentum as both a privacy and cost strategy. Google's SoundStorm 2 Edge demonstrated on-device TTS for mobile in March 2026, Apple's Personal Voice 2.0 runs entirely on-device with Apple Silicon, and Meta's open-source Llama-Voice runs on a single consumer GPU. AMD's entry into the space is significant because it provides an open-source, hardware-agnostic alternative — Lemonade exposes OpenAI-compatible APIs alongside Anthropic and Ollama-style interfaces, meaning developers can swap cloud TTS for local generation with minimal code changes and no vendor lock-in.
Beyond TTS, Lemonade 11.0 ships with a new image-to-3D pipeline via Trellis.2 with an in-app 3D model viewer, ModelScope support as a second remote model registry alongside Hugging Face, a collection router that reports selected routes via response headers, and automatic NPU backend installation on Linux. AMD frames the release as part of a broader vision in which every PC functions as a private AI server capable of text generation, image creation, speech synthesis, and 3D modeling — all running locally. While the TTS and 3D generation features remain labeled experimental, the trajectory is clear: AMD is betting that local, private, open-source AI infrastructure will be a meaningful differentiator as privacy regulations tighten globally and users grow increasingly wary of uploading personal data to cloud services.
