GENERATIVE AI · IMAGE & VIDEO
Generative AI
Generating photorealistic images and video with local AI — from prompt and model to finished result, entirely on my own GPU.
RESULT
Photorealistic images and video, with my own LoRAs trained on a single RTX 4070 — fully local.
01 THE PROBLEM
Good AI images and video aren't a matter of one button. It's in the prompting, choosing the right models, training your own LoRAs and finishing properly. I wanted to master that whole pipeline locally — without a cloud subscription, with full control over style, quality and privacy.
02 WHAT IT DOES
- Photorealistic images
- From portraits and sculptures to cinematic scenes, with attention to light, composition and detail.
- Custom LoRA training
- Fine-tuning models to a specific style or character, trained on my own RTX 4070.
- ComfyUI workflows
- Node-based pipelines for repeatable, controllable results instead of random output.
- Video too
- The same workflow extended to short moving images — from still to clip.
- Fully local
- Everything runs on my own machine: no cloud, no subscription, full control and privacy.
03 WHAT I LEARNED
The difference between ‘a picture’ and a usable result is in the workflow: prompting, iterating, training and finishing. Working locally forced me to understand the whole pipeline instead of leaning on an external service.
INTERESTED?
Let's talk.
Tell me what you think — or what you'd want to do with it.