You need NodeJS and Python3.11 or Docker
Run the following commands for download models and build the project:
bash start.sh
start start.bat
If you want build from sources:
docker-compose build --build-arg SD_MODEL=stabilityai/sdxl-turbo
Then run it:
bash docker.sh
Open your browser at http://localhost:4242
root prompt
, prompt
and size of images
generated at the top right of the screen !
root prompt
is the first template used by ImpAI andprompt
is the template used for every next messages
You can modify the large language model in the file
start.sh
(MODEL
==file name
andMODEL_HF
==repository + quantization
):
# English Mistral 7b Q4_K_M by default
MODEL=mistral-7b-instruct-v0.1.Q4_K_M.gguf
MODEL_HF=TheBloke/Mistral-7B-Instruct-v0.1-GGUF:q4_k_m
# French Mistral 7b Q4_K_M for example
# MODEL=vigostral-7b-chat.Q4_K_M.gguf
# MODEL_HF=TheBloke/Vigostral-7B-Chat-GGUF:q4_k_m
You can change the stable diffusion model (a hugging face repo) in the file
start.sh
# Stable Diffusion Model
SD_MODEL=stabilityai/sdxl-turbo
Chat
Image
Settings
NPC
Save system
Suggested action
Multiple users
RPG mode with dices
Video/GIF output
Audio input