Uses SDXL-Turbo for fast image generation.
Using the RTX 3060 12GB, bot is capable of processing 2 requests per second.
Warning
Note: Model loaded in GPU memory uses ~9.4GB. Can be run on CPU but will require ~2x of memory.
Models remain in memory after a request, for faster processing of subsequent requests.
Memory is freed after an hour of inactivity.
- Install ExApp from AppStore
- Wait for models download and initialization(
6.5
Gb for GPU,13
Gb for CPU) - Enable talk bot in your room
- Use @image mention to send prompts:
@image cinematic portrait of fluffy cat with black eyes
- Image will be posted to room and stored into
image_generator_bot
folder at the user's root.
Important
Only English language is supported
The project is being developed in personal and free time, any ideas or pull requests are welcome.
Note: We understand that the model used here is quite old and cannot compare with the new SDXL-Lightning
We are currently working on a new much more advanced image generation project which you can find here: Visionatrix
If this is of interest to the Nextcloud community, we can try to adapt Visionatrix either as an Image Provider for Nextcloud 30 or simply as an AppAPI Nextcloud application.