
Finally got to training my first model!
The model itself was a heavy time/money/electricity sink, please drop a like if it performed nicely and tips are very welcome so hopefully with some buzz I can accelerate the tagging of the other characters.
It performs extremelly well and its very flexible.
The main trigger words are "Princess Zelda" and "DGMA style"
This is trained on flux D. using Ostris AI toolkit using 38 original high quality AI images made by myself of the character in different clothes settings and poses, initally LLM tagged and then hand polished.
This was a huge effort to do, particularly on the testing to make sure the checkpoint im uploading is the best out of all the outputs I had on the experimental runs.
I have in line 40+ characters ready to go, hopefully I can join the civitai creators program soon.
Please share your creations! Im putting some nice examples myself along with the prompts and settings used to create them. All were created on full precision Flux D. or its Q8 version, I cannot share my Comfy workflow as it includes several custom coded nodes and would expose some of my API keys but am happy to answer questions.
描述:
Initial version
训练词语: Princess Zelda,DGMA Style
名称: Zeldav2_000008000update.safetensors
大小 (KB): 167939
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success