Anime styles are defined by line economy, cel shading, stylized anatomy, and cinematic framing. In AI generation, you guide these levers through prompt anchors and lightweight controls.

Key levers you can steer:

• Line work: thin/clean, brushy, sketchy, variable-width
• Shading: flat cel, two-tone, hard-rim shadows, soft ambient
• Color: pastel, neon, muted filmic, limited palette
• Anatomy/eyes: chibi SD, shojo eyes, shounen proportions, seinen realism
• Backgrounds: painterly, matte-paint, minimalist, graphic gradients
• FX: screentones, speedlines, bloom, rim light, chromatic aberration

Baseline generation tips:

• Aspect ratios: 4:5 or 3:4 for portraits; 16:9 for scenes; 1:1 for icons
• Steps: 20–30 with a modern sampler (e.g., DPM++ 2M)
• CFG/Guidance: 4–7 for stable style adherence
• Resolution: 768–1024 px on the short side; upscale later for detail

Use these anchors to steer toward common anime aesthetics. Mix with subject and setting.

• Classic TV cel (90s): clean lineart, flat cel shading, limited palette, simple backgrounds, retro vibes, film grain subtle
• Modern film look: crisp linework, soft rim light, cinematic color grade, depth of field, high dynamic range
• Shounen action: dynamic pose, motion blur, speedlines, high contrast, bold shadows, energetic composition, debris particles
• Shojo romance: pastel palette, soft bloom, sparkles, delicate eyelashes, floral bokeh, airy lighting
• Seinen gritty/noir: muted colors, heavy shadow, rain, neon accents, film grain, reflective puddles, moody atmosphere
• Mecha/industrial: hard surface detail, panel lines, specular highlights, hangar background, scale cues, smoke haze
• Chibi/super-deformed: big head, tiny body, thick outlines, flat colors, cute props, simplified background
• Fantasy/isekai: painterly background, god rays, ornate costume, magical particles, warm-to-cool gradient
• Cyberpunk neon: rim-lit silhouettes, holograms, dense signage, magenta–cyan palette, volumetric fog
• Monochrome manga: inked lineart, screentones, cross-hatching, black-and-white, panel gutter emphasis

Use these as starting points and swap subject/style anchors.

Character portrait (clean cel):

subject: {hero name}, bust portrait, clean lineart, flat cel shading, two-tone shadows, soft rim light, detailed eyes, subtle gradient background, anime style
camera: 85mm lens, f/2.8, centered composition
color: harmonious limited palette, gentle warm highlights
quality: crisp, high detail, sharp focus

Shounen action scene:

subject: {hero} mid-air attack, dynamic pose, wind-swept clothing, intense expression
style: bold shadows, speedlines, debris particles, impact sparks, high contrast, anime action style
camera: 24mm wide, dutch angle, motion blur, depth of field
environment: ruined street, smoke haze, backlight

Shojo moment:

subject: {two characters} hand-in-hand, soft smiles
style: pastel palette, sparkles, soft bloom, delicate eyelashes, airy lighting, shojo anime style
camera: 50mm, gentle bokeh, center framing

Mecha hangar:

subject: towering mech, maintenance crew, cables and scaffolds
style: hard surface detail, panel lines, specular highlights, cinematic fog
camera: low-angle 28mm, volumetric light shafts

Monochrome manga panel:

subject: {character} dramatic close-up
style: inked lineart, screentones, cross-hatching, black-and-white, bold onomatopoeia
layout: single panel, thick gutters

Negative prompt (drop in as needed):

blurry, extra fingers, deformed hands, messy lineart, muddy colors, low contrast, oversmoothing, text artifacts, watermark, jpeg artifacts

Pose and structure:

• Use pose control (e.g., OpenPose) for strong silhouettes; tweak elbow/knee angles for energy.
• Lineart control: preprocess with Lineart or Canny for crisp inks; keep strength 0.4–0.7 to avoid stiffness.

Composition:

• Rule of thirds for portraits; golden triangle for action.
• Use foreground blockers (smoke, leaves, signage) to add depth.

Camera and lenses:

• Portraits: 50–85mm; action: 20–35mm; mecha scale: 18–28mm low angle.
• Add slight motion blur or radial blur for impact shots; keep readable faces.

Stylization dials:

• Screentones: apply post with halftone filters or texture overlays at 35–55 lpi.
• Speedlines: radial lines behind subject; lower opacity to avoid overpowering lineart.
• Color grade: push warm key light vs cool fill for cinematic contrast.

1. Reference: collect 6–12 images for line weight, shading, palette, and camera.
2. Pose/blocking: set pose control and rough background shapes.
3. Base generation: run low-res draft; lock promising seeds.
4. Style lock: add style anchors (cel/shading/palette); test guidance 4–7.
5. Detail pass: inpaint faces/hands; refine line clarity.
6. Upscale: 1.5–2× with anime-friendly upscaler; optional tile diffusion for textures.
7. Finishing: apply screentones/speedlines, color grade, subtle film grain.

Deliverables to save: prompt, negative prompt, seed, sampler, guidance, Control settings, upscaler.

Common issues and fixes:

• Muddy lineart: increase resolution, use Lineart control at 0.5–0.7, sharpen slightly after upscale.
• Over-rendered shading (not cel): add "flat cel shading, two-tone shadows" and reduce guidance.
• Off-model faces: inpaint face at 512–640 px crop; lower denoise; keep eye highlights consistent.
• Banding in gradients: add subtle grain; use 16-bit workflow when possible.
• Hand artifacts: guide with pose control for hands or inpaint with hand references.
• Neon blowout: cap bloom; lower exposure; add rim light for edge readability.

Which models work best for anime styles? Any modern general model with strong line handling, plus optional anime-focused checkpoints. For tight inks, pair with Lineart control.

Do I need a LoRA for each style? Not always. Start with prompt anchors and control passes. Use LoRA when you need very specific conventions (uniforms, props, or a niche line style).

How do I keep characters consistent across shots? Reuse seed, keep a character descriptor block, fix key traits (hair shape, eye color, accessories), and store reference boards. Consider a lightweight character LoRA if needed.

What resolution should I render? Draft at 768–896 px short side; upscale 1.5–2× for delivery. Upscale after locking composition and face details.