Colab: colab.research...
I was intrigued by a recent opportunity where someone was looking for an AI artist clone to create art in their specific style. Although I didn't get the gig, the challenge fascinated me, and I decided to train an AI to draw in that style. I used an AI model called Dream Booth and Stable Fusion 2.1, and connected it to a new Google Drive account to ensure privacy.
The process involved installing dependencies and creating a model that would take up 2-4GB of space. I provided a link to the collab notebook for others to try. I explained how to install and use a default model to generate images, emphasizing the importance of naming the images properly and running the training for a specific number of steps.
After that, I downloaded a trained model and tested it in a collab file. I opened the checkpoint and generated some art using a specific model. The resulting art style was similar to the style I was aiming for. I experimented with AI-generated images in the style of an artist named Ashwin, starting with a low-resolution image and gradually increasing the number of steps to add more detail. I also used upscaling to increase fidelity and experimented with different prompts, such as wings and Phoenixes.
Join this channel to get access to perks:
/ @mikethetech
Outro music by: Eme7 - "Eme7 x MikeTheTech"
eme7.com/
Discord: exfil.gg/
16 май 2023