My Neuro
An open-source AI desktop companion with local model inference, voice cloning, Live2D characters and long-term memory, licensed under MIT.
My Neuro is a local AI desktop companion for people who want to shape a character's appearance, voice and personality. It pairs replaceable Live2D avatars with spoken conversation, using expressions and movements that respond to the dialogue. The project is open source under the MIT license.
You can run the full inference stack locally, including the language model, or connect a proprietary model through an external API. The local path doesn't require a third-party model API; external model connections send requests to the chosen provider. It supports local model fine-tuning for people who want to adapt the character beyond its appearance.
Voice customization includes cloning and speech-model training through GPT-SoVITS. Speech and subtitles appear together, and you can interrupt the character by voice or keyboard. For characters that speak another language, it can show Chinese subtitles alongside that audio. Typed chat is also available.
Long-term memory retains personal details and preferences, with MemOS providing the memory system. The companion can initiate conversations based on context and use image recognition when a request calls for it. MCP support connects it to tools, including Playwright MCP for browser interaction, while voice commands can open desktop software. Web search requires an internet connection.
For character creators and streamers, My Neuro also supports singing and Bilibili livestreaming. An Android app provides conversations with the bundled Fake Neuro character. Its teaching mode accepts a topic and reference material, and lets you ask questions during a lesson.
