Type | |
Stats | 1,036 964 |
Reviews | (184) |
Published | Aug 26, 2024 |
Base Model | |
Training | Steps: 50 Epochs: 2,000 |
Trigger Words | StoreCCTV |
Hash | AutoV2 5274DFCA5B |
Convenience store CCTV
Day 1/7 of training some proper realism into this plastic AI world.
This is the first Lora I have ever trained.
I'm quite happy with the results so far and I'll keep working on it to learn how to improve LoRas.
It is trained on 450+ frames from different convenience store CCTV recordings as seen on the news.
trigger word: StoreCCTV
strength: 1.5 and over really makes you the cctv security guard
Tested with Flux.1-dev and flux.1-dev-Q8.gguf. The other gguf quantized models should work as well i guess.
reuploaded v1.01 with an exchanged sha256 hash. Seems like ortis' ai-Toolkit generates faulty hashes by generating it from an empty string. Every lora ends up with the same hash, breaking crossposting on civitai (https://github.com/ostris/ai-toolkit/issues/130)
Thanks to diogod for pointing this out! <3