MasaCtrl と SDXL 1.0派生モデルの組合せ結果

こちらはオリジナル記事のサポートページです。

実行結果を載せています。

オリジナル記事

MasaCtrl が一部 SDXL 1.0 (Stable Diffusion XL 1.0) に対応しています。さっそく試してみました。

Realities Edge XL ANIME v1.0

1boy, anime, casual, outdoors, sitting, plain white t-shirt, blue jeans, best quality
1boy, anime, casual, outdoors, standing, plain white t-shirt, blue jeans, best quality
negative_prompt: worst quality, low quality, bad anatomy
seed: 40000

layer44

layer49

layer54

layer59

layer64

Mysterious-SDXL v3.55

prompt1: mysterious, 1boy, anime, casual, outdoors, sitting, plain white t-shirt, blue jeans, best quality
prompt2: mysterious, 1boy, anime, casual, outdoors, standing, plain white t-shirt, blue jeans, best quality
negative_prompt: worst quality, low quality, bad anatomy
seed: 40000

layer44

layer49

layer54

layer59

layer64

haduki_mix v1.0

prompt1: a photo of a pretty japanese woman, outdoors, sitting, plain white t-shirt, blue jeans, best quality
prompt2: a photo of a pretty japanese woman, outdoors, standing, plain white t-shirt, blue jeans, best quality
negative_prompt: worst quality, low quality, bad anatomy
seed: 30000

layer44

layer49

CounterfeitXL v1.0

prompt1: 1boy, anime, casual, outdoors, sitting, plain white t-shirt, blue jeans, best quality
prompt2: 1boy, anime, casual, outdoors, standing, plain white t-shirt, blue jeans, best quality
negative_prompt: worst quality, low quality, bad anatomy
seed: 20000

layer44

layer49

layer54

layer59

layer64

LastpieceCoreXL s0964

prompt1: 1girl, kawaii, anime, outdoors, sitting, plain white t-shirt, blue jeans, best quality
prompt2: 1girl, kawaii, anime, outdoors, standing, plain white t-shirt, blue jeans, best quality
negative_prompt: worst quality, low quality, bad anatomy
seed: 30000

layer44

BreakDomainXL v03d

prompt1: 1girl, kawaii, anime, outdoors, sitting, plain white t-shirt, blue jeans, best quality
prompt2: 1girl, kawaii, anime, outdoors, standing, plain white t-shirt, blue jeans, best quality
negative_prompt: worst quality, low quality, bad anatomy
seed: 30000

layer59

layer64

Realities Edge XL v2.0

prompt1: a photo of a pretty woman, outdoors, sitting, plain white t-shirt, blue jeans, best quality
prompt2: a photo of a pretty woman, outdoors, standing, plain white t-shirt, blue jeans, best quality
negative_prompt: worst quality, low quality, bad anatomy
seed: 30000

layer59

layer64

コメント

コメントを投稿

このブログの人気の投稿

【AnimateDiff】Motion Module v2 を使った動画作成の結果

イメージ
こちらはオリジナル記事のサポートページです。 実行結果を載せています。 オリジナル記事 【AnimateDiff】Motion Module v2 が公開されて作成動画の質が向上したそうです。さっそく試してみました。 お気に入りの結果（yabalMixTrue25D_v4） AnimateDiffで作成したのは16フレームです。 FPS 4にして4秒動画にしています。 model: yabalMixTrue25D_v4.safetensors seed: 1000000000 steps; 40 guidance_scale: 7.5 prompt: a girl, dancing, blue denim, white plain t-shirt, best quality, extremely detailed negative_prompt: longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality guidance_scaleを変えると結果はどう変わるか？ 左からguidance_scaleを6.5→7.5→8.5に変化させた場合の結果です。 toonyou_beta6 model: toonyou_beta6.safetensors seed: 2000000000 steps; 40 guidance_scale: 8.5 prompt: a girl, dancing, blue denim, white plain t-shirt, best quality, extremely detailed negative_prompt: longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality upscaling: GFPGAN CounterfeitV30_v30 model: CounterfeitV30_v30.safetensors
続きを読む

fuduki_mix と haduki_mix の画像比較

イメージ
左が「fuduki_mix」、右が「haduki_mix」です。 ストリートに立つ女性（fuduki_mix v1.0 vs haduki_mix v1.0） prompt: portrait of young japanese girl, 25yo, 8k, detailed, standing on street, smiling, plain white t-shirt, eye level angle negative_prompt: None 25歳（fuduki_mix v1.0 vs haduki_mix v1.0） (Compel) prompt = "beautiful japanese woman with smile, 25yo+++, 8k, RAW photo, (best quality)+, masterpiece, photo-realistic, focus, professional lighting" negative_prompt = "(worst quality)++, (low quality)++" seed: 10000, 20000, 30000, 40000, 50000 no refiner 35歳（fuduki_mix v1.0 vs haduki_mix v1.0） (Compel) prompt = "beautiful japanese woman with smile, 35yo+++, 8k, RAW photo, (best quality)+, masterpiece, photo-realistic, focus, professional lighting" negative_prompt = "(worst quality)++, (low quality)++" seed: 10000, 20000, 30000, 40000, 50000 no refiner 45歳（fuduki_mix v1.0 vs haduki_mix v1.0） (Compel) prompt = "beautiful japanese woman with smile, 45yo+++,
続きを読む

ControledAnimateDiff ( ControlNet + AnimateDiff ) の結果

イメージ
こちらはオリジナル記事のサポートページです。 実行結果を載せています。 オリジナル記事 ControlNet と AnimateDiff を組み合わせて Video2Video に挑戦 conditioning_scale: 0.75 model: toonyou_beta6.safetensors motion_module: mm_sd_v15.ckpt control: conditioning_scale: 0.75 controlnet_processor: softedge controlnet_processor: control_v11p_sd15_softedge guess_mode: true seed: 1000000000, 2000000000, 3000000000, 4000000000 steps: 40 guidance_scale: 7.5 prompt: a girl, dancing, blue denim, white plain t-shirt, best quality, extremely detailed n_prompt: longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality conditioning_scale: 0.85 conditioning_scale: 0.95 conditioning_scale: 1.0
続きを読む