https://github.com/iamaziz/llm-voice-bot

Speak (speech-to-text) to Ollama LLMs in any lanaguage - Streamlit app
https://github.com/iamaziz/llm-voice-bot

llms ollama streamlit voice-bot voice-recognition

Last synced: 4 days ago
JSON representation

Speak (speech-to-text) to Ollama LLMs in any lanaguage - Streamlit app

• Host: GitHub
• URL: https://github.com/iamaziz/llm-voice-bot
• Owner: iamaziz
• License: mit
• Created: 2024-02-27T04:20:22.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-02-27T20:09:20.000Z (9 months ago)
• Last Synced: 2024-10-12T04:53:30.155Z (about 1 month ago)
• Topics: llms, ollama, streamlit, voice-bot, voice-recognition
• Language: Python
• Homepage:
• Size: 35.2 MB
• Stars: 33
• Watchers: 2
• Forks: 8
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# 🎙️ Voice LLM Bot

Speak to Ollama LLMs in any language.

## Demo (📢 voice on)

https://github.com/iamaziz/llm-voice-bot/assets/3298308/e444d4db-f8d9-4b7d-996d-cb783d704c61



**Note**

- If a langauge is not available in the dropdown list, it can be added [here](https://github.com/iamaziz/llm-voice-bot/blob/main/app.py#L11).

- Thanks to [streamlit_mic_recorder](https://github.com/B4PT0R/streamlit-mic-recorder).



> Interesting how LLMs responded differently to switching languages in the same conversation.

> Almost all the tested models understood the question in its original langauge, but they varied in how they reponded.

> Mistral models consistently answered in English. While other models stumbled going back and forth.

Here are some examples:

mistral

![image](https://github.com/iamaziz/llm-voice-bot/assets/3298308/4d0b97c9-e42e-4d83-9cd2-fee970cba86f)

dolphin-mistral



llama2:13b



qwen:7b



gemma:7b (worst)



And GPT-4 for comparison: