Updated to V2
The tags have been changed.
The cape area has been improved a bit. Mainly around the chest.
By separating the tags, I think the feathers on the hat are now printed more stably.
Other than that, I don't think there's much difference.
I'm grateful to everyone who gave me advice.
タグを変更しました。
ケープ周りが少し改善されました。主に胸当たり。
タグ分けして、安定して帽子の羽根は出力されるようになったと思います。
それ以外はそこまで変わりはないかと思います。
アドバイスをくれた方に感謝です。
This LoRA recreates the elegant Red Mage costume inspired by Final Fantasy V.
It features a feathered red hat, a gold-trimmed red cape, and a one-piece dress, capturing the classic and charming look of a magical swordswoman.
Final Fantasy V 風の赤魔導士衣装を再現したLoRAです。 赤い帽子に羽根飾り、金の縁取りが特徴的なケープとワンピース風ドレスを学習しています。 クラシックかつ可憐な魔法剣士スタイルを再現したい方におすすめです。
Strength:0.8~1.2
Costume Prompts:
red_mage_cos,red headwear,hat feather,red cape,red dress,capelet,gold trim,belt,thigh bootsDescription
FAQ
Comments (5)
Works nicely. I note "Red Mage (Final Fantasy)" works surprisingly well on its own as a tag, but it's such a mix of disparate ideas it's good to have LoRAs for particular versions.
I'm ashamed to say that I didn't generate images using the "Red Mage (Final Fantasy)" tag to check what kind of results I could get (lol).
But I think it's good that it has been established as one style of Red Mage.
@tiwaz10 it's better that you didn't, using the "Red Mage (Final Fantasy)" tag is using latent knowledge, which generally is less flexible than a lora should be.
It does the job, it's a bit style burned, with a preference for low angle shots. If you were to retrain this, i'd drop the learn rate a bit, and remove 'red feathered hat' for 'red headwear' and 'hat feather', so you don't have two custom tags competing for absorbing information in your lora.
For people looking to use this for male characters, or who just prefer pants, this prompt should get the job done with the lora, no custom tag needed:
red headwear, hat feather,red cape, red tunic, gold trim, belt, knee boots, red pants, capeletCheers.
Thanks for the advice.
I think there was probably one low-angle shot in the training image that was strongly reflected.
I'll split the tags into two, lower the learning rate, and retrain it.
Oh, a pants style prompt!
Amazing!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.