I tested it a few times, but I feel like it doesn't capture the face very accurately. Especially for some sensitive characters, if you change the face a little, the customer won't like the work... In this case, what should be added or subtracted to make it look completely similar?
Lora is the answer. But you need at least 5 high quality photos of a person. If you have just one, it is some combination of pullid, instantid and reactor. Maybe generate sdxl image then img2img in flux to refine. There is no simple solution. For some faces one works better than the other.
![SahBabii - Viking [Official Music Video]](http://i.ytimg.com/vi/D37cL92JX0I/mqdefault.jpg)
![[SFM] Risky Recital](http://i.ytimg.com/vi/RzNkY1_Nk3o/mqdefault.jpg)
its about time to try pullid
here since many days 😁
i been tryin to sort this shit out all DAY man. watched ur vid 2hrs after u posted and gonna neck. fkn insightface python aids
@@donblaa are you having issues with insight face?
@@AIFuzz59 yea was, fixed it. only took me 10 hours
How to write a text in a U-Nai Get Text? I have it empty(
I tested it a few times, but I feel like it doesn't capture the face very accurately. Especially for some sensitive characters, if you change the face a little, the customer won't like the work... In this case, what should be added or subtracted to make it look completely similar?
Lora is the answer. But you need at least 5 high quality photos of a person. If you have just one, it is some combination of pullid, instantid and reactor. Maybe generate sdxl image then img2img in flux to refine. There is no simple solution. For some faces one works better than the other.
Its been available for a couple months… it takes too much resources and nit great results
hibernation maybe