Trained on a new dataset that was hand captioned.

1. Use Natural langue with a trigger word or try your own.

2. Can use at 1.0 without degradation of quality in most cases.

3. Tested up-scaling 0.5-0.7 at 1536x1536

Thank you for continuing to support my training addiction with your tip (By default I know) of just 1 buzz per generation while using this lora.

An example of my cost:

1. 11¢ per hour for training power consumption

2. 15¢ per hour for amortization of equipment cost over 3 years

3. For this lora for my charged time on the dataset would have been $200