Assinatura PixAI
PixAI
PixAI
Generate
Sign in
Página Inicial
Geração
Modelos
Estúdio
NEW
Concurso
Criar com Mio.2
Caixa de Ferramentas
NEW
Documento PixAI

LoRA de Usuário

DiT.2
Estilo

(説明必読・特殊条件)ドット絵小型化style  pixel art

#Pixel art

Criar

  • Gerador de Imagens com IA
  • Gerador de Animação com IA
  • Caixa de Ferramentas
  • Geradores Temáticos
  • Treinar LoRA
  • Agente Mio.2
  • Estúdio

Descobrir

  • Tags de Tendências
  • Classificação
  • Mercado de Modelos
  • Concurso
  • Notícias

Sobre

  • Documento PixAI
  • Como Usar a PixAI
  • Tsubaki.2
  • Conheça a Mio
  • Regras de Conteúdo

Preços e Ajuda

  • Assinatura
  • Pacotes de Créditos
  • Contato

Aplicativo Móvel

  • Baixe o app PixAI
  • Loja de Aplicativos
  • Google Play
© 2026 PixAI
  • Termos de Serviço
  • Política de Privacidade
  • Política de Direitos Autorais
  • Lei de Transações Comerciais Especificadas
(説明必読・特殊条件)ドット絵小型化style  pixel art - AI Model cover image
Treinado por
Avatar image of 七夢
avatar-frame

七夢

Treine Sua Própria LoRA
Enviado em
20 de jun. de 2026, 02:47
Usos
368
Obtido
20,105
Avaliações
Sem avaliações
Palavras Gatilho

late_super_famicom_tactical_rpg_tiny_battle_character_sprite_with_visible_pixels_hard_black_outline_restricted_16bit_palette_simplified_anatomy_flat_shading_small_eyes_minimal_facial_features_and_low_screen_occupancy,

Permissões
  • Permitir geração e compartilhamento de imagens
  • Permitir que usuários baixem seu modelo
  • Usos comerciais

Descrição

1.はじめに ・本styleLoRAはキャラをドット絵+小型化させ、ファミコン・スーパーファミコン風のキャラクターを生成することが目的です。

・どのようにして遊ぶのかについてはイラストを複数投稿して実例を挙げてゆきます。

・うごイラ「SFC風テニスゲーム」のような遊び方も可能です。

・興味があり、遊んでみようかな?という場合には、下記をご確認頂いた上で、ご使用いただければと思います。

2.遊び方

⭐テストの結果、一番安定した生成ができるのではないか?と思う設定です。これ以外でも遊べない訳ではありませんが、成功率は落ちると感じています。

・LoRAの強さは1.0以上が推奨です。

・モードはプロが推奨です。(ネガティブを使用するので) スタンダードでも遊べないこともないのですが、成功率は落ちます。

・サイズは1024×1024が推奨です。

・慣れるまでは一括×4での生成がおすすめです。

・LoRAを併用する場合、併用LoRAとの相性がすべてです。

オリジナルトリガー+キャラLoRAのトリガーでも遊べますが、キャラの全身が分かるようなプロンプトでない場合には、必ずキャラの全身が分かるプロンプトを追加してください。 (例)「Take a full-body image that shows all the way down to your feet.」

ネガティブプロンプトを1度削除して、以下をネガティブプロンプトに追加してください。 「large_character, close_up, portrait, full_size_character, detailed_full_body_illustration, modern_anime, smooth_gradient, high_detail, super_deformed, chibi, big_head,」

⭐併用する場合のLoRAの強さについて。

・本LoRA 1.0、キャラLoRA 0.7 で生成してみて、ドット絵+小型化しないな?と感じたら、キャラLoRAは0.2以下に引き下げて下さい。

・他のスタイルLoRAを併用する場合には、同じ比率に設定(1.0)し、状況を見ながら強さを調整ください。

⭐あれ?上手くいかない、又は、より小型化させたいと感じたら以下を試してみてください。

・本LoRAの強さを上げてください。

・以下のプロンプトを、トリガーワードの後に追加してください。 「wide shot, distant camera, zoomed out view, character far from camera, tiny character in the center, low screen occupancy, large background space, full environment visible」

⭐背景が描かれてこないと感じたら。

・以下のネガティブプロンプトをさらに追加してください。 「close up, portrait, upper body, large character, huge character, full size character, character fills frame, centered large character, detailed full body illustration, poster composition, hero shot, modern anime illustration」

扱いずらいLoRAですが、少しでも楽しいな~と感じて頂ければ、嬉しいです♪

(Please read the explanation carefully - Special conditions) Pixel Art Miniaturization Style

  1. Introduction
  • This style of LoRA aims to generate characters in a pixel art + miniaturization style, resembling characters from the Famicom/Super Famicom era.

  • I will post several illustrations to illustrate how to play.

  • It's also possible to play in a similar way to the animated illustration "SFC-style tennis game."

  • If you're interested and want to try it, please check the following before using it.

  1. How to Play

⭐ Based on testing, these settings seem to produce the most stable generation. While it's not impossible to play with other settings, the success rate will likely decrease.

  • A LoRA strength of 1.0 or higher is recommended.

  • The Professional mode is recommended (because it uses negatives).

  • It's not impossible to play in Standard mode, but the success rate will decrease.

  • A size of 1024 x 1024 is recommended.

  • Until you get used to it, generating in batches of 4 is recommended.

  • When using LoRA in conjunction, compatibility with the other LoRA is everything.

You can play with the original trigger + character LoRA trigger, but if the prompt doesn't show the character's full body, be sure to add a prompt that shows the character's full body.

(Example) "Take a full-body image that shows all the way down to your feet."

Delete the negative prompt once, and add the following to the negative prompts:

"large_character, close_up, portrait, full_size_character, detailed_full_body_illustration, modern_anime, smooth_gradient, high_detail, super_deformed, chibi, big_head,"

⭐Regarding the strength of LoRA when using in conjunction:

  • If you generate with LoRA 1.0 and character LoRA 0.7 and feel that the pixel art + miniaturization isn't working, lower the character LoRA to 0.2 or lower.

  • When using other LoRA styles, set them to the same ratio (1.0) and adjust the strength while observing the results.

  • If it's not working, or you want to make it smaller, try the following:

  • Increase the strength of this LoRA.

  • Add the following prompt after the trigger word: "wide shot, distant camera, zoomed out view, character far from camera, tiny character in the center, low screen occupancy, large background space, full environment visible"

  • If you feel the background isn't being drawn:

  • Add the following negative prompt: "close up, portrait, upper body, large character, huge character, full size character, character fills frame, centered large character, detailed full body illustration, poster composition, hero shot, modern anime illustration"

LoRA can be tricky to use, but I hope you find it even a little bit fun!

Comentários