https://github.com/eryk-mazus/sigh

Seamless Voice Interactions with LLMs
https://github.com/eryk-mazus/sigh

llm speech-recognition speech-to-text voice-recognition whisper

Last synced: 2 months ago
JSON representation

Seamless Voice Interactions with LLMs

• Host: GitHub
• URL: https://github.com/eryk-mazus/sigh
• Owner: eryk-mazus
• Created: 2023-09-02T09:39:45.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-10-28T12:50:36.000Z (about 1 year ago)
• Last Synced: 2024-07-30T21:03:13.914Z (5 months ago)
• Topics: llm, speech-recognition, speech-to-text, voice-recognition, whisper
• Language: Python
• Homepage:
• Size: 69.3 KB
• Stars: 11
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# sigh

Seamless Voice Interactions with LLMs

**Key Features:**

* **Unlimited Real-time Transcription:** Continuously capture audio directly from your microphone.

* **Customizable Wake Word:** Choose a wake word or phrase to trigger transcription mode.

* **Automatic Speech Termination:** Detects when you've finished speaking, with an option for manual control.

**Note:** This repository is under active development. Contributions are welcome!

**Demo:**

## Setup:

```

set OPENAI_API_KEY=sk-...

git clone https://github.com/eryk-mazus/sigh.git

cd sigh

pip install -e .

# run:

python ./sigh/main.py --help

# run without wake word detection (by default):

python ./sigh/main.py

# run with wake phrase detection:

python ./sigh/main.py --detect_wake_phrase=True --wake_phrase="""Hey GPT"""

```

## Backlog:

Near-term:

- [x] Add automatic transcription stopping

- [x] Better GPT responses (system prompt, chat mode, sliding memory buffer)

- [ ] Talk with local models, e.g. llama2, mistral, etc.

- [ ] Improve code coherence and composition (refactoring)

Medium-term:

- [ ] Add second mode: parallel transcription and LLM commentary

- [ ] Docker

## Contributing:

Issues, new ideas, suggestions, and PRs are all welcome!