RepoPicks Radar

Updated every day around 23:30 UTC

Guides

Open-LLM-VTuber

Open-LLM-VTuber/Open-LLM-VTuberReviewed by AI

Checked by RepoPicks Radar on

What is Open-LLM-VTuber?

Open-LLM-VTuber is an AI companion you talk to out loud. A Live2D cartoon character answers with a cloned or synthetic voice, shows expressions, can be interrupted mid-sentence and can sit on your desktop as a transparent pet window. You choose the parts: a local model through Ollama or LM Studio, or a paid cloud API; several speech-recognition and text-to-speech engines. The Windows and macOS downloads are the desktop client only — the server behind it is installed with Python, uv and ffmpeg, or with Docker, and configured in a conf.yaml file.

Open-LLM-VTuber is free and open source (MIT) and can replace ChatGPT and ElevenLabs.

How easy is it to install Open-LLM-VTuber?

D · terminal, several tools or a specific graphics card · Windows

  • Windows D
  • macOS D
  • Linux D

All three are D. The releases only hold the desktop client (open-llm-vtuber-1.2.1-setup.exe, 96 MB, and the 165 MB .dmg); the part that does the work is a Python server you install with uv plus ffmpeg, or the included Dockerfile, then edit conf.yaml to pick a language model, a speech-recognition engine and a voice. Linux has no download at all and the README warns that reaching the server from another device needs https set up behind a reverse proxy.

Coming soon: install in 1 click with the RepoPicks app.

Open-LLM-VTuber on GitHub

Stars
14,667on Nov 10, 2026
Downloads
42,390496 this week
Growth
+83stars in 6 days

Stars over time

14,62014,63014,64014,65014,66014,670 Nov 7Nov 8Nov 9Nov 10 14,62014,63014,64014,65014,66014,670 Nov 7Nov 9

Facts

Runs on
Windows, macOS, Linux
You need to install
Python, Docker, FFmpeg, NVIDIA graphics card
To build from source
Python, Docker, NVIDIA graphics card, FFmpeg
Disk space
96 MB–289 MB · from the installer
Accounts or keys
OpenAI or any OpenAI-compatible service, Anthropic (Claude), Azure Speech (recognition and voices), ElevenLabs, Cartesia or Fish Audio voices, Groq
Cost
Free · MIT licensed, except the included Live2D sample characters. It can run entirely on local models at no cost; the cloud options it supports (OpenAI, Claude, Azure Speech, ElevenLabs, Fish Audio, Groq) are paid per use.
License
MIT (Permissive)
Last change on GitHub
May 15, 2026

Open-LLM-VTuber risk signals (6)

  • AttentionDependencies with known vulnerabilities

    Pinned dependencies with known vulnerabilities (OSV): aiohttp: GHSA-2fqr-mr3j-6wp8, GHSA-2g87-pp6c-29x5, GHSA-2vrm-gr82-f7m5; brotli: GHSA-2qfp-q593-8484, GHSA-9v8h-qqgm-4m35, PYSEC-2026-2401; mcp: GHSA-5h93-6whr-6q8j, GHSA-84m7-p3x7-pcfv, GHSA-9h52-p55h-vw2f; requests: GHSA-gc5v-m9x4-r6x2, PYSEC-2026-2275; torch: GHSA-3749-ghw9-m3mg, GHSA-53q9-r3pm-6pq6, GHSA-63cw-57p8-fm3p; torch: GHSA-63cw-57p8-fm3p, GHSA-qfhq-4f3w-5fph, GHSA-rrmf-rvhw-rf47; aiohttp: GHSA-2fqr-mr3j-6wp8, GHSA-2g87-pp6c-29x5, GHSA-2vrm-gr82-f7m5; anyio: GHSA-5p39-cfhj-2xmp, GHSA-82r6-8w77-94w6, PYSEC-2026-4024; certifi: GHSA-6q48-6c7w-9gw2, PYSEC-2026-4320; cryptography: GHSA-537c-gmf6-5ccf, GHSA-g6cj-pr64-35w5, GHSA-jwv3-5hgf-82ww; dnspython: GHSA-45c3-73f7-pv4m; filelock: GHSA-69hj-f68j-2879, GHSA-8mx5-7m5m-r77g; fsspec: GHSA-27vj-qcqg-25rc, PYSEC-2026-4217; idna: GHSA-65pc-fj4g-8rjx, PYSEC-2026-215; mcp: GHSA-5h93-6whr-6q8j, GHSA-84m7-p3x7-pcfv, GHSA-hvrp-rf83-w775; multidict: GHSA-54p9-h82j-f925, PYSEC-2026-4244; protobuf: GHSA-7gcm-g887-7qv7, PYSEC-2026-1805; pydantic-settings: GHSA-4xgf-cpjx-pc3j; pygments: GHSA-5239-wwwm-4pmq, PYSEC-2026-2987; pyjwt: GHSA-2gx3-rcp4-g85q, GHSA-42vr-xj54-vc7v, GHSA-752w-5fwx-jx9f; python-dotenv: GHSA-mf9w-mj56-hr94, PYSEC-2026-2270; python-multipart: GHSA-5rvq-cxj2-64vf, GHSA-6jv3-5f52-599m, GHSA-mj87-hwqh-73pj; requests: GHSA-gc5v-m9x4-r6x2, PYSEC-2026-2275; setuptools: GHSA-h35f-9h28-mq5c, GHSA-xmj3-9gf7-h52w, PYSEC-2026-3447; soupsieve: GHSA-2wc2-fm75-p42x, GHSA-836r-79rf-4m37, GHSA-gjv8-xp57-g29c; starlette: GHSA-82w8-qh3p-5jfq, GHSA-86qp-5c8j-p5mr, GHSA-jp82-jpqv-5vv3; torch: GHSA-3749-ghw9-m3mg, GHSA-53q9-r3pm-6pq6, GHSA-63cw-57p8-fm3p; torch: GHSA-rrmf-rvhw-rf47, PYSEC-2026-139; urllib3: GHSA-8988-9cw3-xx77, GHSA-gh4c-6fx4-qh6g, GHSA-mf9v-mfxr-j63j; virtualenv: GHSA-394h-qwmc-48pf, GHSA-43mp-35q3-x3fq, GHSA-494p-63j8-28p3

    automatic check

  • Attentionit can watch your camera and screen

    README feature list: "Visual perception, supporting camera, screen recording and screenshots, allowing your AI companion to see you and your screen". If you configure a cloud model instead of a local one, those pictures of you and your screen are sent to that company.

    AI review

  • Attentionincluded characters are not under the mit licence

    LICENSE and the README: the Live2D sample models "are licensed separately under the Live2D Free Material License Agreement ... They are not covered by the MIT license of this project", and "For commercial use, especially by medium or large-scale enterprises, the use of these Live2D sample models may be subject to additional licensing requirements."

    AI review

  • Infoversion 1 is on hold for a rewrite

    README: "We are focusing on Open-LLM-VTuber v2.0 — a complete rewrite of the codebase ... We kindly ask you to refrain from opening new issues or pull requests for feature requests on v1." It also warns "This project is in its early stages". The last release is v1.2.1 from 2025-08-26 and the last commit is 2026-05-15.

    AI review

  • Infodownloads models outside the project folder

    The uninstall section says models fetched "via ModelScope or Hugging Face may also be in MODELSCOPE_CACHE or HF_HOME" rather than the project's models directory, so removing the app can leave gigabytes behind.

    AI review

  • Infosold as a virtual girlfriend or boyfriend

    README: "whether you want a virtual girlfriend, boyfriend, cute pet, or any other character, it can meet your expectations", the repository topics include ai-waifu and ai-companion, and the User Reviews section quotes "This girlfriend has been used over 100,000 times". The persona is a prompt you write, so what the character says depends on the model and prompt you choose.

    AI review

Evidence is shown in its original language.

What Radar could not check about Open-LLM-VTuber

  • Whether the voice conversation really keeps up on a computer without a graphics card
  • Total disk space once speech, voice and language models are downloaded
  • Whether the installed desktop client talks to the server without extra fiddling on each system
  • What the cloud providers keep of your voice, camera and screen images if you choose them
  • Whether the many listed speech and voice engines all still work in version 1.2.1

Similar projects

The RepoPicks app

Coming soon: install in 1 click with the RepoPicks app.

Before you download

See the release page on GitHub