Sumitc13

/

Flux-Kontext-exposure-control-LoRAs

like 6

Hi,
First of all thank you for sharing your ev brackets LoRAs.
I've been using them for a while and they do a great job.
I would like to support this initiative and propose you to use the same dataset to train LoRAs for Flux.Klein.9B.base and Qwen.Edit.2511.
I wanted to know if you would be interested and if you would accept my help to do so.
Cheers, and thanks again for sharing.
Alberto

Hi Alberto,
Good to know you found these useful. I felt that this process is way too slow and not that effective, I have better methods now to generate HDR, so I am not currently training these for newer models. If you using it for something else or some other method you found to merge these to HDR, I would love to know..

Hi!
Thanks for the prompt answer. I agree with you that requiring 4 inferences to get the 4 ev brackets takes time but with qwen-edit-2511 and the lighting LoRA or flux.klein.9b distilled this inference time should be reduced, or at least I would love to validate that hiphothesis.
Regarding the generation of HDR from the ev brackets, I ended up using a version of what Photoshop uses. It's true that is not perfect but is plausible enough.

I was wondering if you would be able to share the training data, so I can carry on training the LoRAs for other models and sharing it here for whoever wants to use them.

I'm also very interested in the alternative improved ways to generate HDR from DLR images. Anything you could share? 🤤

Hi!
Thanks for the prompt answer. I agree with you that requiring 4 inferences to get the 4 ev brackets takes time but with qwen-edit-2511 and the lighting LoRA or flux.klein.9b distilled this inference time should be reduced, or at least I would love to validate that hiphothesis.
Regarding the generation of HDR from the ev brackets, I ended up using a version of what Photoshop uses. It's true that is not perfect but is plausible enough.

I was wondering if you would be able to share the training data, so I can carry on training the LoRAs for other models and sharing it here for whoever wants to use them.

I'm also very interested in the alternative improved ways to generate HDR from DLR images. Anything you could share? 🤤

Cool. I'll try to train the loras for klein 9B. My computer is busy training some other models. I'll do this as soon it's free..

Hi!
Thanks for the prompt answer. I agree with you that requiring 4 inferences to get the 4 ev brackets takes time but with qwen-edit-2511 and the lighting LoRA or flux.klein.9b distilled this inference time should be reduced, or at least I would love to validate that hiphothesis.
Regarding the generation of HDR from the ev brackets, I ended up using a version of what Photoshop uses. It's true that is not perfect but is plausible enough.

I was wondering if you would be able to share the training data, so I can carry on training the LoRAs for other models and sharing it here for whoever wants to use them.

I'm also very interested in the alternative improved ways to generate HDR from DLR images. Anything you could share? 🤤

Cool. I'll try to train the loras for klein 9B. My computer is busy training some other models. I'll do this as soon it's free..

Happy to help tranning LoRAs as well with my computer, if you send me the recipe I will follow the steps