Sign up free, get 10,000 credits
Starter credits added when you sign up.
working on an eye LoRA for facial expressions,
working on an eye LoRA for facial expressions, AI art by Cin&Cia on PixAI. Made with PixAI Edit Pro.
**Key Things to Consider When Making a LoRA** 1. **Prepare and carefully select materials that match your target style.** More images don’t necessarily mean better results. Your dataset still needs to be curated, otherwise you may run into data contamination and tolerance issues. 2. **Anything you don’t provide data for is left for the model to interpret.** Information that isn’t clearly represented in the training data will often be left for the model to decide or fill in during generation. 3. **Write captions to tell the model what’s in each image.** The model learns based on what you write in the captions. If you leave them blank, you’re letting the model interpret the images on its own. But if your captions are wrong, those mistakes may also be learned and lead to incorrect results later. (This stage is usually great for working with AI. Depending on the task, it can speed things up by around 8–20×.) 4. **Sometimes, choosing NOT to caption something is just as important.** More detailed captions aren’t always better. What you label and what you intentionally leave out should depend on what you actually want the LoRA to learn. So before sending it into the oven, ask yourself: **“What do I actually want this LoRA to express?”** --- **LoRAを作るときに意識したいポイント** 1. **目標とするスタイルに合った素材を用意し、しっかり選別する。** 画像は多ければ多いほど良いというわけではありません。データはきちんと選別しないと、データの汚染や許容範囲の問題につながる可能性があります。 2. **データとして与えていない部分は、モデル自身の解釈に任せることになる。** 学習データの中で明確に与えられていない情報は、実際に画像を生成するとき、モデル側の判断や補完に委ねられることになります。 3. **Captionを書いて、その画像に何があるのかをモデルに伝える。** モデルはCaptionに書かれた内容をもとに学習します。書かなければモデル自身の判断に任せることになりますが、間違った内容を書けば、その間違いまで学習され、後の生成結果に影響する可能性があります。 (この工程はAIとの共同作業と相性がよく、内容によっては作業速度を8〜20倍ほどまで上げられます。) 4. **あえてCaptionに書かない、という選択が必要なこともある。** Captionは詳しければ詳しいほど良い、というものではありません。何を書くのか、そして何をあえて書かないのか。それは、このLoRAに何を学ばせたいのかによって変わります。 だから、オーブンに入れる前に一度考えてみる。 **「このLoRAに、本当に表現してほしいものは何だろう?」** --- **製作 LoRA 時需要注意的幾個要點** 1. **準備並篩選符合目標風格的素材。** 素材並不是越多越好。資料本身仍需要篩選,否則可能產生資料汙染,也需要考慮模型的容錯問題。 2. **沒有提供資料的部分,就留給模型自行發揮。** 訓練資料中沒有明確提供的資訊,在之後生成圖片時,往往就會交由模型自行判斷與補足。 3. **撰寫 Caption,告訴模型圖片裡有什麼。** 模型會根據 Caption 的內容進行學習。如果不寫,就等於讓它自行判斷;但如果 Caption 寫錯,錯誤的資訊也可能被一起學進去,進而影響日後的生成結果。 (這個階段通常很適合與 AI 協作,依照工作內容不同,效率大約可以提升 8~20 倍。) 4. **有些東西,反而要選擇不寫進 Caption。** Caption 並不是寫得越詳細就一定越好。哪些資訊要標註、哪些資訊要刻意留下,其實都取決於你究竟想讓這顆 LoRA 學會什麼。 所以在送進烤箱以前,還是得先想清楚: **「我真正想讓這顆 LoRA 表現出來的東西,到底是什麼?」** --- **LoRA를 만들 때 고려해야 할 몇 가지 포인트** 1. **목표로 하는 스타일에 맞는 자료를 준비하고 신중하게 선별하기.** 이미지는 많다고 해서 무조건 좋은 것은 아닙니다. 데이터는 반드시 선별하는 과정이 필요하며, 그렇지 않으면 데이터 오염이나 모델의 허용 범위와 관련된 문제가 생길 수 있습니다. 2. **데이터로 제공하지 않은 부분은 모델의 해석에 맡겨지게 된다.** 학습 데이터에서 명확하게 제공하지 않은 정보는 이후 이미지를 생성할 때 모델이 스스로 판단하고 채워 넣게 됩니다. 3. **Caption을 작성해 이미지에 무엇이 있는지 모델에게 알려 주기.** 모델은 Caption에 작성된 내용을 바탕으로 학습합니다. 작성하지 않으면 모델이 스스로 이미지를 판단하게 되지만, 잘못된 Caption을 작성하면 그 오류까지 함께 학습되어 이후 생성 결과에 영향을 줄 수 있습니다. (이 과정은 AI와 협업하기 좋은 부분이며, 작업에 따라 속도를 약 8~20배까지 높일 수 있습니다.) 4. **때로는 일부러 Caption에 적지 않는 것도 중요하다.** Caption은 자세하게 쓸수록 무조건 좋은 것이 아닙니다. 무엇을 적고 무엇을 의도적으로 남겨 둘지는 결국 이 LoRA가 무엇을 배우길 원하는지에 따라 달라집니다. 그러니 오븐에 넣기 전에 한 번 생각해 봅니다. **“나는 이 LoRA가 정말 무엇을 표현하길 원하는가?”**
Similar Images
Show moreあは
m0
((((pochi_(pochi-goya)))), furina_(genshin_impact), the character has long, flowing hair that is primarily white with light blue highlights and streaks. Her eyes are blue, and she has an ahoge. She is wearing a light blue dress shirt with a dark blue bow-tie or ascot, and the hem of the shirt is showing over white denim shorts. She is barefoot. The character is sitting cross-legged on a gaming chair with armrests, and she is hugging her knees with her hands around her ankles. She is smiling directly at the viewer from a first-person perspective from a computer monitor, as if she is livestreaming. The room is at night with a side desk in the background. The image is a fake screenshot with a livestream overlay, including a chat log scrolling from top to bottom in a black area on the left side of the screen, and a seek bar in a black area at the bottom of the screen. The style features the artist's signature clean delicate linework, voluptuous yet stylized bishoujo anatomy, soft cel-shading with smooth skin gradients, luminous bloom effects, dramatic rim lighting, cool-toned dominant palette, and expressive glossy eyes. The atmosphere is vibrant and intimate.
mamamyamya#0
m0
((((pochi_(pochi-goya)))), furina_(genshin_impact), the character has long, flowing hair that is primarily white with light blue highlights and streaks. Her eyes are blue, and she has an ahoge. She is wearing a light blue dress shirt with a dark blue bow-tie or ascot, and the hem of the shirt is showing over white denim shorts. She is barefoot. The character is sitting cross-legged on a gaming chair with armrests, and she is hugging her knees with her hands around her ankles. She is smiling directly at the viewer from a first-person perspective from a computer monitor, as if she is livestreaming. The room is at night with a side desk in the background. The image is a fake screenshot with a livestream overlay, including a chat log scrolling from top to bottom in a black area on the left side of the screen, and a seek bar in a black area at the bottom of the screen. The style features the artist's signature clean delicate linework, voluptuous yet stylized bishoujo anatomy, soft cel-shading with smooth skin gradients, luminous bloom effects, dramatic rim lighting, cool-toned dominant palette, and expressive glossy eyes. The atmosphere is vibrant and intimate.
mamamyamya#0
あは
Explore more
Recommended Creators
You might like
Enchanting Mouse-Eared Beauty
Zombii
11
🍑⸝꙳.˖
「 🎤 𖥔 . 𝘢𝘪𝘳𝘪 ♡ ! 💕 」
𓏧 ♡ 👑 🌎
「 🎤 𖥔 . 𝘢𝘪𝘳𝘪 ♡ ! 💕 」
。゚🎂 ✧ 🍦 ⋆.
「 🎤 𖥔 . 𝘢𝘪𝘳𝘪 ♡ ! 💕 」
A beautifully rendered anime illustration of a blue-violet-haired girl reclining sideways in an elegant, antique rose-brown armchair inside a luxurious, softly illuminated room. She rests her cheek against one hand, her elbow supported by the chair's armrest, with a quiet, slightly melancholic expression and half-lidded eyes gazing toward the viewer. Her body curls naturally into the seat, while her bent legs and flowing layered skirt extend diagonally toward the lower-left foreground.
She wears an elaborate black-and-lavender gothic dress with puffed sleeves, a fitted bodice, delicate white lace, layered ruffles, and a pale decorative apron, complemented by a wide-brimmed hat adorned with ribbons and flowers. The palette blends dusty rose, muted mauve, lavender, cool blue-gray, ivory, and deep charcoal.
Soft lavender-pink lighting streams through the enormous decorative windows, creating luminous highlights along her hair, hat, sleeves, lace, and the polished wooden floor. Gentle violet shadows define the folds of her dress and the contours of the plush armchair, while warm reflected light adds depth to the carved wooden frame. The intricate baroque interior features ornate floral carvings, curved window frames, delicate flower arrangements, and polished parquet flooring.
A dramatic tilted perspective and low, close camera angle emphasize the chair's sweeping silhouette and the character's relaxed pose. Fine anime linework, smooth gradient shading, subtle ambient occlusion, delicate fabric textures, and painterly light reflections create a sophisticated, dreamy atmosphere. The vertical composition combines intricate gothic elegance with soft pastel lighting and richly detailed environmental rendering., painterly digital rendering, chunky brushwork, broad opaque strokes, faceted shapes, angular color blocking, hard-edged shading, painterly edges, soft tonal transitions, diffused highlights, ambient lighting, cinematic lighting, subtle bloom, textured illustration, grainy texture, crayon texture, chalky brushwork, impasto strokes, bold color blocking, stylized anime, flat shading, dry brush effect, atmospheric lighting, soft glow highlights, blocky-like illustration, loose textured pencil linework, crayon-like lineart, flat minimal shading, dry-brush textured transitions, chalky stippling, bold saturated tonal palette, high-contrast color blocks
Rosetta
7
character: kaname_madoka, akemi_homura, A warm, romantic anime-style illustration of Madoka and Homura depicted in an intimate, fairytale-like scene, composed as though viewed through an elaborate antique vanity mirror.
Madoka is positioned in the center foreground, facing the viewer directly while sitting in front of the mirror. Her long, straight dark-brown hair falls over her shoulders, framing her face with soft bangs and a bright red ribbon tied near the top of her head. Her large violet eyes are open, with a gentle, slightly surprised expression and parted lips. She wears a delicate white ruffled dress with an off-shoulder neckline, layered fabric, small bows, and a high white choker.
Behind her, Homura appears within the mirror's reflection, standing close and gently arranging Madoka's hair. Homura has long, flowing pinkish-purple hair tied back with a pale ribbon, with loose strands framing her face. Her eyes are closed and she wears a peaceful, affectionate expression. She is dressed in a flowing white dress with delicate ruffled sleeves and soft fabric detailing. One hand delicately holds and arranges Madoka's hair while the other extends outward, holding a section of red ribbon, as though carefully styling her hairstyle.
Composition and setting: The central focus is a large, vertically oriented antique mirror with an ornate golden frame and a softly curved upper edge. Its reflection contains Homura and Madoka, creating a layered composition in which Madoka occupies the lower foreground and Homura appears slightly elevated behind her. The mirror is surrounded by rich crimson-red wallpaper decorated with intricate repeating floral patterns.
On the left side, a tall white vase holds an arrangement of lush red roses, smaller cream-colored flowers, and green foliage. Along the bottom foreground is an elegantly arranged tea table with delicate porcelain teacups and saucers, a small slice of cake, and a pale green tablecloth. These details reinforce the intimate, old-fashioned, romantic atmosphere.
Rendering and coloring: Soft, painterly anime illustration with warm ambient lighting, delicate brushwork, and subtle textured shading. The artwork combines clean anime facial features with loosely painted surroundings. Hair is rendered with layered, individually suggested strands and broad highlights, while the dresses feature soft folds, translucent-looking fabric, and delicate cream-colored highlights.
The palette is dominated by deep burgundy, antique crimson, warm gold, dusty pink, chocolate brown, ivory, and muted sage green. The red wallpaper and roses create a rich backdrop, contrasted by the luminous white dresses and golden mirror frame.
Lighting is especially warm and atmospheric, with golden illumination falling across the characters' faces, hair, and clothing. Soft shadows around the mirror frame and background give the scene depth, while subtle highlights make the fabric and hair appear luminous.
The overall aesthetic is elegant, nostalgic, intimate, and fairytale-like, with a vintage Victorian tea-room atmosphere, romantic character interaction, ornate decorative framing, and warm, painterly anime rendering., loose energetic brushwork visible textured brush strokes rough sketch like linework thin dark outlines mixed with soft painted edge, simplified yet expressive facial features, soft blended rendering, strong highlights and strong shadows, dramatic lighting, painterly digital rendering, chunky brushwork, broad opaque strokes, faceted shapes, angular color blocking, hard-edged shading, painterly edges, soft tonal transitions, diffused highlights, ambient lighting, cinematic lighting, subtle bloom, semi-realistic, textured paint layering, delicate lines, impasto brushwork, smooth shading, glowing highlights, muted chromatic tones, subtle vivid accents, cinematic grain, sparkling irises, dreamlike atmosphere
Rosetta
2
This elaborate, painterly anime-style illustration presents a dark-fantasy portrait of a female character with vibrant, textured brushwork and a rich color palette dominated by deep greens, blacks, and glowing golden accents. The subject has messy, shoulder-length chartreuse green hair with loose strands framing her face, and piercing, luminous yellow-gold eyes that gaze forward with a calm, solemn expression. She wears a massive, wide-brimmed black witch's hat adorned with intricate textured patterns along the crown and decorated with a cluster of pale white lily flowers tucked into the side, from which a sheer black veil or fabric cascades downwards. Her upper body is clad in a sophisticated, Victorian-inspired dark blouse or gown featuring a high ruffled white collar, dark layered ruffles on the sleeves, and a structured bodice fastened with dark buttons and a dark brooch at the throat. In her hands, she gently clutches a bouquet of bright white lilies with delicate green stems, while ethereal glowing lights, sparkles, and faint floral or vine-like elements drift around her in the ambient space. The background consists of an ornate, dark architectural frame or window with golden highlights and a moody stained-glass or paneled effect, and a watermark reading "COMMISION @旋转熊猫" is subtly overlaid across the lower portion of the composition., painterly digital rendering, chunky brushwork, broad opaque strokes, faceted shapes, angular color blocking, hard-edged shading, painterly edges, soft tonal transitions, diffused highlights, ambient lighting, cinematic lighting, subtle bloom, textured illustration, grainy texture, crayon texture, chalky brushwork, impasto strokes, bold color blocking, stylized anime, flat shading, dry brush effect, atmospheric lighting, soft glow highlights, semi-realistic anime illustration, painterly digital rendering, loose textured brushwork, soft blended shading, ethereal glow, sketchy linework, dramatic colored lighting, subtle cross-hatching, selective saturated accents
Rosetta
3
artist: Rourow201310207, artistic, masterpiece, cinematic, ultra_detailed:1.6, no hard outlines, ambiental light, textured brushwork, lively color_art, Beautiful colors, art style, 4k. A dramatic, darkly romantic anime illustration depicting an elegant woman reclining diagonally against a richly saturated red backdrop, surrounded by deep crimson roses and scattered foliage. The composition is inspired by gothic fashion photography, with an intimate, theatrical atmosphere and a striking interplay between warm illumination and velvety shadows.
Composition and pose: The woman occupies most of the composition, positioned diagonally across the frame with her head resting toward the upper-left area and her body extending toward the lower-right. The viewpoint is slightly elevated, looking down at her reclining figure. Her posture is relaxed and expressive, with her upper body gently twisted toward the viewer. Her head tilts slightly sideways as she gazes directly toward the viewer with a languid, mysterious expression. Her eyes are half-lidded, framed by exceptionally long, dramatic eyelashes and dark eyeliner. Her lips are softly parted and richly defined. Her voluminous, nearly black hair spreads loosely around her head, creating an irregular, flowing silhouette against the red background. Several loose strands escape around her face and extend outward, while the remaining hair forms thick, tousled waves. One arm extends toward the lower-left side, holding a slender rose stem between her fingers. Her other hand rests across her waist, delicately gathering several long rose stems. Her fingers are relaxed and elegantly positioned.
Clothing and surrounding details: She wears a fitted black dress with a delicate lace-trimmed neckline and fine ornamental detailing. The fabric is predominantly dark and absorbs much of the surrounding light, with subtle highlights revealing its folds, texture, and fitted silhouette. A large bouquet of deep red roses rests across her chest and torso. Several fully opened flowers overlap one another, accompanied by dark green leaves and slender stems that extend down across her body. Additional roses and foliage appear along the lower-left edge, partially disappearing into the shadows.
Background and atmosphere: The entire setting is enveloped in a rich, uninterrupted crimson-red environment. The background is minimal, without distracting architectural elements, allowing the flowers, figure, and dramatic lighting to dominate. The roses and leaves create delicate organic shadows against the red surface, adding depth and reinforcing the romantic gothic atmosphere.
Rendering style and lighting: A sophisticated, high-detail anime illustration with a semi-realistic painterly finish. Refined anime facial proportions, expressive eyes, delicate linework, and carefully blended shading. The skin has a luminous, warm appearance, rendered with smooth gradients and subtle transitions between golden highlights and reddish-brown shadows. Softly illuminated cheeks, delicate nose shading, glossy lips, carefully defined eyelids, and individual eyelashes. Hair in dense, overlapping locks with loose wispy strands, subtle reflective accents, and deep, almost-black internal shadows. The black dress has restrained satin-like reflections, subtle fabric folds, and delicate lace details. The roses are richly dimensional, with individually layered petals, curled edges, soft highlights, and deep shadows. The leaves have slightly reflective surfaces and delicate vein details.
Lighting: Intense, directional warm lighting with a strong amber-golden illumination falling across the woman's face, neck, shoulders, arms, and hands. Concentrated and theatrical, creating a golden glow against the saturated crimson environment. Opposite sides of her body, her dark clothing, and portions of her hair disappear into deep burgundy and near-black shadows. Subtle red reflected light along the edges of her hair, dress, and arms. Dramatic chiaroscuro, strong shadow separation, soft atmospheric falloff, and gentle highlights across the rose petals. The background remains intensely red but slightly darker toward the outer edges, creating a vignette-like effect.
The overall rendering is rich, soft, and painterly. Dark romanticism, gothic floral fashion editorial, luxurious crimson atmosphere, dramatic chiaroscuro, warm amber illumination, semi-realistic anime painting, glossy facial details, velvety black fabrics, richly layered roses, and cinematic low-key lighting.
Rosetta
5
character: vivian_banshee, copyright: zenless_zone_zero
vivian banshee, zenless zone zero, 1girl, solo, long hair, looking at viewer, shirt, long sleeves, holding, closed mouth, red eyes, meds, ahoge, purple hair, outdoors, pa, frills, hairband, sky, pointy ears, mole, night, drill hair, umbrella, frilled sleeves, black nails, building, night sky, twin drills, holding umbrella, cityscape, skyscraper, purple umbrella, semi-realistic anime style, spiky eyelashes, glossy eyes, textured digital painting, fine cross-hatching, scratchy pencil-like linework, visible canvas grain, detailed screen tones, chromatic aberration, intense under-lighting, glowing rim light, high-contrast values, luminous skin, beautiful light, best shadow, color blink, impasto, painterly rendering, thick refined brushstrokes, visible paint texture, glossy eyes, delicate eyelashes, luminous glow, painterly hair texture, reflective hair highlights, smooth gradients, soft bloom, atmospheric glow, gentle facial features, fluid hair movement, elegant composition, diffused highlights, subtle rim lighting, cinematic light scattering
Rosetta
Sg75
1
cyber
𝚖𝚎𝚛𝚒𝚖𝚎𝚛𝚒
...
Tofu
『秋のケモ耳愛護週間🍁#5』
JOA
🐈⬛🐈
ゆゆ
❄️
ゆゆ
さ
夏の陽光とサマードレス
さば味噌煮鰆
LoRA Test 1💚💛
Haiyo💝
embers
ᯓᯓ 𝙘𝙤𝙘𝙤𝙛𝙡𝙪𝙛𝙛𝙞
나비 소녀
조개
etc.
綾戯 ゆべし
character: hatsune_miku, This iconic anime illustration, created for the Vocaloid song "Deep Sea Girl" (深海少女), features Hatsune Miku submerged in a dark, atmospheric underwater world.
Positioned in a dynamic perspective from above, Miku floats upward through deep ocean waters, reaching her right hand directly toward the viewer with delicate, pale fingers extended in a gentle, yearning gesture. Her signature bright teal twin-tails fan out expansively behind her, floating weightlessly across the blue background alongside soft, glowing water droplets and bubbles that reflect light. She has wide, expressive dark eyes glistening with tears and a slightly parted mouth, conveying a poignant mix of sadness and hope.
She wears her classic black "Deep Sea Girl" dress, shown here in deep shadow with silhouetted ruffles blending into the dark water around her lower body, contrasting sharply against the bright cyan of her hair and the soft light hitting her skin and face. On the right side of the composition, white Japanese typography reading "深海少女" along with the English subtitle "- deep sea girl -" is cleanly set against the dark blue backdrop., painterly digital rendering, chunky brushwork, broad opaque strokes, faceted shapes, angular color blocking, hard-edged shading, painterly edges, soft tonal transitions, diffused highlights, ambient lighting, cinematic lighting, subtle bloom, yamiguro style, painterly, rough visible brush strokes, impasto texture, heavy chiaroscuro, extreme contrast, harsh lighting, limited color palette, psychological horror, unsettling atmosphere, grainy texture, film grain, rough painterly anime illustration, expressive textured brushwork, chunky brushstrokes, chalky dry-brush texture, scraped pigment, blocky paint application, soft blended shading, visible paint texture, sketchy painted details, glossy highlights, atmospheric depth
Rosetta
🍓
utka
the fairy from the lake
ᯓᯓ 𝙘𝙤𝙘𝙤𝙛𝙡𝙪𝙛𝙛𝙞
🐑⛰️🎻
♡▪︎ ꩇׁׅ݊υׁׅꭈׁׅᧁׁ℘ꪱׁׅᝯׁhׁׅ֮ꪱׁׅ℘ᨮ꫶ׁׅ֮~🍀▪︎♡
12
Comments