- Control panel that allows you to skip unwanted requests.
- Custom characters with custom personalities and custom voices. (you can create your AI persona)
- Uses state of the art TTS and best role playing LLM models.
- Word filter List. You can filter words you don't want to listen or see, both from the chatter requested and from the AI responding.
- Different types of fun interactions like plain AI request, AI talking to each other, TTS, and forsen-style TTS with soundbites and filters.
- Obs script to add skip keybind.
- Forsen forsen forsen forsen forsen forsen forsen forsen forsen forsen.
- Security features (added control panel sharing and fixed filters, so I guess it's done)
- Add images to chars
- Tune LLM
- Obiwan voice
- Switched to Index-TTS
- AI talking to each other
- Integrate OpenMemory in all AI related flows
- Investigate ways to reduce VRAM footprint of index-tts-vllm
- TTS Streaming Support to reduce first word latency
- Fix show image not working sometimes and not working when request is not active yet
- Fix skip button stopping working
- Make better images guide
- Show response when state of prompt image changes and show image visibility state
- Move all static files(images and voice references) from postgres to minio
- Full switch from production version to beta
- Refactor fundamental flaw in applying background filters. They only work on single audio/sfx right now. But they should be applied on full concatenated audio until filter is popped. Same goes for left to right/right to left filter.
- Fix AI agent unable to stop conversation in Agent BAJ reward.