lllyasviel
Problem lifting the prompt from PNG via Image browser to txt2img
This is what Image Browser shows:
Prompt: Extreme closeup hyperrealistic photo of a bald J0hfr3 man <lora:j0hfr3_man_Flux.dev.lora_000004000:1> with a full beard as a realistic version of Homer Simpson from the Simsons TV-show, movie still, <lora:xlabs_flux_realism_lora_comfui:1> <lora:ugly_flux:0.5>
Lora hashes: j0hfr3_man_Flux.dev.lora_000004000: e3b0c44298fc, xlabs_flux_realism_lora_comfui: 8c6dd764ceff, ugly_flux: 5a4becfcc624
This is what gets read in to txt2img:
Extreme closeup hyperrealistic photo of a bald J0hfr3 man <lora:lynkcorim_lora:1> with a full beard as a realistic version of Homer Simpson from the Simsons TV-show, movie still, <lora:xlabs_flux_realism_lora_comfui:1> <lora:ugly_flux:0.5>
It changes to the lynkcorim-lora and it's not even in the folder of lora's any more. 🤔
Possible solution: In Settings -> Infotext -> Disregard fields from pasted infotext, add Lora hashes to the list.
I'm seeing the same issue on my end too, it seems to swap the LoRAs to random ones despite the metadata having the correct ones written down. The solution @DenOfEquity suggested solved it for me!
Lora hashes
There's no Lora hashes for me to select.
Update: Found a new fix for myself. Probably it was changed in newer version: It's now a separated option called "When adding to prompt, refer to Lora by" You have to pick "Filename" instead of "Alias from file".
