[ AI Art ] - stable diffusion 討論 分享 02
武吉士
704 回覆
9 Like
8 Dislike
第 1 頁第 2 頁第 3 頁第 4 頁第 5 頁第 6 頁第 7 頁第 8 頁第 9 頁第 10 頁第 11 頁第 12 頁第 13 頁第 14 頁第 15 頁第 16 頁第 17 頁第 18 頁第 19 頁第 20 頁第 21 頁第 22 頁第 23 頁第 24 頁第 25 頁第 26 頁第 27 頁第 28 頁第 29 頁
1.5*
主體係oldfish
唔記得有冇溝其他model
唔記得有冇溝其他model
即係我係咪要放d 有角嘅相落去train, 咁喺其他相，例如貓貓狗狗嘅相就會得出有角嘅結果
所以你練果時要俾埋tag 佢，等佢唔好咁易train 到所有野都有角。
Civital 上面成日有啲LoRA 亂鳩咁落tag又overtrained ， 好易會搞到貓貓狗狗都變埋獨角獸。
Civital 上面成日有啲LoRA 亂鳩咁落tag又overtrained ， 好易會搞到貓貓狗狗都變埋獨角獸。
用下mov2mov 整動畫, 唔落ebsynth似乎都係好難解決flickening 問題...用multiple control net 又有bugs....
have u seen my test?
Ebsynth is much better than mov2mov
Ebsynth is much better than mov2mov
想問下見到D AI KOL佢地可以整到個臉張張差唔多樣係點整出黎
定係佢地都係用明星既lora合成整到差不多樣
定係佢地都係用明星既lora合成整到差不多樣
ebsynth 用過,只係懶D 想係webui 一野過輸出 , 純粹試下, 見D大陸人玩mov2mov 玩得好勁....
This is mov2mov, but not as bad as yours.
其實唔難 
未出Ebsynth Utility果陣玩過一陣放棄左, 依家有懶人包試下先.....
未試Ebsynth Utility,我用嗰個係另外一個.
Ebsynth有個致命問題係每次keyframe最多20張,要人手分batch做, 同埋佢個ui好廢
等緊佢出Ebsynth Studio 1.0
Ebsynth有個致命問題係每次keyframe最多20張,要人手分batch做, 同埋佢個ui好廢
等緊佢出Ebsynth Studio 1.0
咩setting可以咁consistent
multicontrolnet + denoise唔好太高
controlnet我用咗 canny + depth
controlnet我用咗 canny + depth
咁要爬下YouTube
我明明地，UI 唔friendly 唔記得點整
我明明地，UI 唔friendly 唔記得點整
no prompts?
this is my prompts 
with some basic negative prompts....
korean girlwith some basic negative prompts....
Denoise set得唔高其實冇咗mov to anime嘅意義。
見啲日本仔gen出黎幾好睇喎， 試吓先
理論上freeze (masked)左個background , 落兩個control net ,denoise 可set 高啲，再用ebsynth blend 理一齊應該真係做到animation 效果，不過未試到，local 機得6GB vRam 整幾野就爆
yes, but judge by result, not the denoise value,
set the denoise value range I think 0.2-0.5
to a point that the image is change to the style you want of course.
set the denoise value range I think 0.2-0.5
to a point that the image is change to the style you want of course.
有冇D prompts 可以減少係咁閃
，件衫都係咁轉
，件衫都係咁轉