Use a combination of ChatGPT and VoiceVox to make cool chat robots.
- Install VoiceVox Engine. I recommend using the pre-made docker images.
- (If using VoiceVox for Japanese voices): Make the VoiceVox Engine available on port 50021. If you're using docker, you can do this with
docker run -p 50021:50021 -it [DOCKER CONTAINER YOU DOWNLOADED]. - Sign up for the chatgpt api, and save your API key.
Before beginning, we just need to set two environment variables inside a .env file.
OPENAI_API_KEY=[Your Api Key Here]
VOICEVOX_API_URL=[Your VoiceVox Endpoint Here]
If you don't plan to use VoiceVox and just want a streaming chat gpt service, you can leave the VOICEVOX_API_URL variable unset.
When run without any arguments, chat-stream will work approximately as ChatGPT does. It will take messages from the command line and return chatGPT responses in a streaming mode.
Commands:
--vv Use VoiceVox to read responses out loud using a preset voice
--vv-api Run the program as an api server in VoiceVox mode. See Api Doc below.
--cid [Character ID] Use a specific VoiceVox character
An informal list of supported characters and corresponding ids can be found here.
Audio samples and licensing for those characters can be found here.
When run with the --Api flag, chat-stream will run as an api server.
The server has one endpoint.
| Method | Endpoint | Header | Body | Response |
|---|---|---|---|---|
| Post | /get-voice | contentType-json | {voice_lines: string} | audio/wav |