AI Chat Generator | Chat AI with Google Flan-T5
Chat AI including a free chatbot demo
Due to popular demand I have released the Chat AI experimental chatbot as a separate free demo. The full version includes the ability to experiment with the full power of the T5-Flan models. Both features will be getting updates in the near future.
Files
AI Chat Generator | Chat AI with Google Flan-T5
AI Text like ChatGPT on your own PC
| Status | Canceled |
| Category | Tool |
| Author | Capsize Games |
| Tags | ai,artificial-intelligence,bot,chat-bot,Generator,Open Source,tool |
| Languages | German,English,Spanish; Latin America,French |
More posts
Comments
Log in with itch.io to leave a comment.
| Updated | Feb 29, 2025 |
| Published | Mar 23, 2026 |
| Status | Ordered Purchased |
| Category | Tool |
| Rating | Rated 4.3 out of 5 stars(6total ratings) |
| Author | Capsize Games |
| Tags | ai, artificial-intelligence, bot, chat-bot, Generator, Open Source, tool |
| Average session | About a half-hour |
| Languages | German, English, Spanish; Latin America, French |
| Inputs | Keyboard, Mouse |
No - you should avoid buying if the demo does not work.
If you don’t mind:
- what video card do you have?
- what os did you test on?
If you are comfortable with code, you can also try running from source:
https://github.com/Capsize-Games/chatai
You can also try downloading and running AI Runner which is free and built with the same architecture:
https://capsizegames.itch.io/ai-runner
If AI Runner works, then Chat AI might also work.
Your card might not be able to handle the flan-t5-xl which is what the chatbot is using. You can either wait a future build or try to install the open source version anf make the chatbot use a smaller model.
Without more details it’s hard to say what is going wrong on your end.
I suspect you might be able to run airunner but not Chat AI as the memory management for diffusers models is better than the flan model.
I will be releasing an updated chatbot (including the demo) soon that allows switching to smaller models, and I’m looking into LLaMas which can run in 4bit mode. Those things should help lower powered cards, so keep an eye on this page.