Vellium v1.1.0 — Live voice, local STT/TTS and easier llama.cpp setup
热度趋势
趋势数据积累中
百分比基于当前可用热度信号，而非评论数或独立用户人数。
Vellium v1.1.0 是一款开源的本地优先桌面应用程序，专为 AI 聊天、角色扮演和长篇写作而设计。它将聊天记录、角色、LoreBooks、写作项目和知识集合本地存储在 SQLite 中。Vellium 支持 macOS、Windows 和 Linux 操作系统，并兼容 OpenAI-compatible APIs、OpenRouter、LM Studio、Ollama 和 KoboldCpp。开发人员正在积极征求用户反馈，特别是关于本地语音或角色扮演设置中任何感到不便或过于复杂之处。
Vellium is an open-source, local-first desktop app for AI chat, character roleplay and long-form writing.
Recent updates have focused mostly on making local voice and model setups easier to use.
Live mode now supports microphone input, local or Whisper-compatible speech recognition, streaming TTS, attachments, screen context and the usual chat tools—all inside the same voice interface.
Local speech can be installed and configured directly in the app. Whisper Large v3 Turbo Q5_0 is available for recognition, while TeraTTSv2 provides English and Russian voices with realtime playback. The TTS process stays active between responses, avoiding a full model reload for every reply.
The llama.cpp setup has also been simplified. Vellium can detect existing llama-server installations, GGUF models and running local endpoints, then configure them as a managed backend.
There have been plenty of smaller fixes as well: more reliable TTS streaming, safer runtime archive extraction, better endpoint discovery, improved timeout handling, system certificate support and easier settings navigation.
Chats, characters, LoreBooks, writing projects and knowledge collections are stored locally in SQLite. Vellium runs on macOS, Windows and Linux and supports OpenAI-compatible APIs, OpenRouter, LM Studio, Ollama and KoboldCpp.
Feedback from people using local voice or roleplay setups would be especially useful-particularly about anything that still feels awkward or unnecessarily complicated.