This repository has the following 3 features:
- Conversation with AI character (ChatVRM feature)
- AITuber streaming
- WebSocket mode
I've written a detailed usage guide in the article below:
- Install packages
npm install
- Start the application in development mode
npm run dev
- Open the URL http://localhost:3000
- This is a feature to converse with an AI character.
- It is a feature of pixiv/ChatVRM, which is the basis of this repository.
- It can be tried relatively easily as long as you have an API key for various LLMs.
- The last 10 conversation sentences are retained as memory. (The number can be specified in future updates)
- Enter your API key for various LLMs in the settings screen.
- OpenAI
- Anthropic
- Groq
- Ollama (No API key is required, but a local API server needs to be running.)
- Edit the character's setting prompt if necessary.
- Load a VRM file if needed.
- Select a speech synthesis engine and configure voice settings if necessary.
- For VOICEVOX, you can select a speaker from multiple options. The VOICEVOX app needs to be running beforehand.
- For Koeiromap, you can finely adjust the voice. An API key is required.
- For Google TTS, languages other than Japanese can also be selected. Credential information is required.
- For Style-Bert-VITS2, a local API server needs to be running.
- Start conversing with the character from the input form. Microphone input is also possible.
- It is possible to retrieve YouTube streaming comments and have the character speak.
- A YouTube API key is required.
- Comments starting with '#' are not read. (The string can be specified in future updates)
- Turn on YouTube mode in the settings screen.
- Enter your YouTube API key and YouTube Live ID.
- Configure other settings the same way as "Conversation with AI Character".
- Start streaming on YouTube and confirm that the character reacts to comments.
- You can send messages to the server app via WebSocket and get a response.
- Unlike the above two, it does not complete within the front-end app, so the difficulty level is a bit higher.
- Start the server app and open the
ws://127.0.0.1:8000/ws
endpoint. - Turn on WebSocket mode in the settings screen.
- Configure other settings the same way as "Conversation with AI Character".
- Wait for messages from the server app and confirm that the character reacts.
- You can try it with the server app repository I created. tegnike/nike-open-interpreter
- For detailed settings, please read "Let's develop with a beautiful girl!! [Open Interpreter]".
- The license adheres to pixiv/ChatVRM.
- Language settings support Japanese and English. You can switch in the settings screen.
- Conversation history can be reset in the settings screen.