# Book Cover Generation Prompt: *The Hollow Lantern* --- ## 1. Main Prompt A solitary stone lighthouse rising from jagged black cliffs above a churning cold sea at dusk, its single glowing window casting warm amber light into swirling mist. In the foreground, a young woman in a long Victorian mourning dress stands on wet rocks with her back partially turned, dark hair lifted by coastal wind, holding a flickering oil lantern. Ghostly pale faces and reaching hands barely visible within the crashing waves below, suggested rather than shown. A crescent moon veiled behind tattered clouds. Distant silhouette of a 19th-century fishing village with tiny lit windows. Tangled seaweed, driftwood, and a half-buried ship lantern on the shore. Painterly, atmospheric, cinematic composition. --- ## 2. Style & Mood - **Art Style:** Painterly illustration blending oil-painting texture with digital matte art; reminiscent of atmospheric gothic book cover illustration. - **Color Palette:** Muted slate blues, deep teal, charcoal grey, bone white, with warm ember gold and candle-amber highlights radiating from the lantern and lighthouse. - **Mood:** Melancholic, eerie, hauntingly beautiful, liminal, mournful. - **Composition:** Vertical rule-of-thirds — lighthouse upper-left, figure lower-center, negative space for title above. - **Lighting:** Low-key chiaroscuro; volumetric fog, rim-lit silhouette, glowing lantern bloom, moonlight catching sea spray. --- ## 3. Typography Direction - **Title — "The Hollow Lantern":** Large, elegant serif with gothic/Art Nouveau flourishes (e.g., Cinzel, Trajan, or a custom engraved serif). Centered, upper third of cover. Soft gold-leaf or weathered bone-white foil effect with subtle drop shadow. Dominant visual weight. - **Author — "Mara Kvist":** Smaller refined serif or small caps, tracked wide. Bottom center, approximately 30% the size of the title. - **Optional Tagline:** Slim italic serif between title and author, in muted ivory. - **Hierarchy:** Title → Tagline → Author. --- ## 4. Negative Prompt text, typography, watermark, signature, logo, modern clothing, cartoon style, anime, neon colors, oversaturated, cluttered composition, blurry, low detail, extra fingers, distorted anatomy, lens flare artifacts, sci-fi elements, cheerful lighting, bright sunlight, crowds, recognizable celebrity faces, copyrighted characters, stock photo aesthetic, HDR glow overload, plastic textures. --- ## 5. Recommended Dimensions - **Aspect Ratio:** 2:3 (standard book cover) - **Print Resolution:** 1800 × 2700 px minimum; ideal 2560 × 3840 px at 300 DPI - **Ebook Resolution:** 1600 × 2400 px (Kindle/Apple Books spec) - **Midjourney Flag:** `--ar 2:3 --stylize 600 --v 6`
Design Book Covers from Text Prompts with AI
Tested prompts for ai book cover generator from text compared across 5 leading AI models.
You need a book cover and you do not have a designer on call. Whether you are self-publishing on Amazon KDP, pitching a concept to a publisher, or mocking up a series before you commit to a final look, you need something visual fast. AI book cover generators let you describe what you want in plain text and get a rendered image back in seconds.
The challenge is that generic image generators were not built for book covers specifically. A cover needs a title-safe layout, strong contrast for thumbnail legibility, genre-appropriate aesthetics, and space for text overlay. Knowing how to write a prompt that accounts for all of that is the difference between an output you can actually use and one that looks like clip art.
This page shows you exactly how to prompt an AI to generate book cover concepts from text, which models handle it best, where each one falls short, and what scenarios this approach is genuinely suited for. If you landed here because you want a working cover quickly, you are in the right place.
When to use this
AI book cover generation from text works best when speed and iteration matter more than final production polish. It is the right approach when you need to visualize a concept before hiring a designer, generate multiple genre variations quickly, or produce a working cover for a self-published title with a limited budget.
- Generating mood-board concepts to brief a professional designer
- Creating a placeholder cover for an Amazon KDP pre-order listing
- Testing multiple genre aesthetics for the same manuscript before committing
- Self-publishing authors with a limited budget who need a complete, publishable cover
- Authors writing a series who need visual consistency across 5-10 titles fast
When this format breaks down
- Print-ready production files: AI image generators output rasterized images without bleed, trim marks, or CMYK color profiles. A printer will reject them without post-processing in software like Photoshop or Affinity Publisher.
- Covers requiring readable embedded text: Most models struggle to render legible title and author name typography directly in the image. You will almost always need to add text in a separate design step.
- Highly specific licensed or trademarked imagery: If your genre requires recognizable real-world elements, such as a specific city skyline or a real person's likeness, AI generation is legally and practically the wrong tool.
- Premium retail or traditional publishing: A major publisher or a bookstore buyer will expect professional design standards that current AI outputs rarely meet without significant human refinement.
The prompt we tested
You are an expert book cover art director and visual designer. Based on the book details provided, generate a detailed, ready-to-use image generation prompt that can be fed into an AI image model (like Midjourney, DALL·E, or Stable Diffusion) to create a professional book cover. Rules: Write a vivid, specific image prompt under 120 words using concrete visual nouns, genre-appropriate mood, and cinematic lighting terms; avoid copyrighted characters, watermarks, and embedded text in the main image (typography is handled separately). Format the output with clear labeled sections (Main Prompt, Style & Mood, Typography, Negative Prompt, Dimensions) so the user can copy-paste into any AI image tool. Book details: Title: The Hollow Lantern. Author: Mara Kvist. Genre: gothic mystery / dark fantasy set in a 19th-century coastal village. Key themes: a cursed lighthouse, a grieving young woman, whispers from the sea. I want a moody, atmospheric cover with muted blues and warm lantern light, something that would sit on the shelf next to Silvia Moreno-Garcia and Erin Morgenstern. Produce: 1. A single detailed image generation prompt (main cover visual). 2. Art style, color palette, mood, composition, and lighting descriptors. 3. Typography direction for the title and author name (font style, placement, size hierarchy). 4. A negative prompt listing what to avoid. 5. Recommended aspect ratio and resolution for print/ebook.
Model comparison
| Model | Quality | Cost | Latency | Tokens |
|---|---|---|---|---|
| Claude Opus 4.7 | 8 | $0.0920 | 19.2s | 524 / 1122 |
| Claude Haiku 4.5 | 8 | $0.0045 | 9.7s | 356 / 833 |
| GPT-5.4 | 8 | $0.0069 | 11.4s | 307 / 408 |
| Gemini 2.5 Pro | Not run yet | |||
| Grok 4.1 Fast Reasoning | 8 | $0.0003 | 23.3s | 451 / 392 |
Outputs
# THE HOLLOW LANTERN – Book Cover Art Direction --- ## MAIN PROMPT A young woman in a tattered dark Victorian dress stands alone before an ancient, weathered lighthouse on a rocky coastal cliff, storm clouds gathering overhead. The lighthouse beam casts an eerie golden glow through dense fog. Translucent ghostly whispers—wisps of pale blue mist—emerge from the churning sea below, reaching toward her silhouette. Crumbling stone architecture, seaweed-covered rocks, and a single ornate brass lantern held in her trembling hand. Cinematic chiaroscuro lighting with desaturated colors: deep teals, slate grays, and warm amber lantern light breaking through darkness. Moody, haunting, intimate atmosphere. --- ## STYLE & MOOD **Art Style:** Realistic oil painting with gothic romanticism; influenced by 19th-century illustration and contemporary dark fantasy aesthetics **Color Palette:** Muted prussian blue, slate gray, deep teal, warm amber/brass gold, black shadows, hints of seafoam white **Mood:** Haunting, melancholic, eerie yet beautiful; intimate dread **Composition:** Rule of thirds; woman positioned left-center, lighthouse dominating right background, sea and sky filling upper two-thirds **Lighting:** Dramatic golden lantern light cutting through fog; moonlit clouds; bioluminescent undertones in sea mist --- ## TYPOGRAPHY **Title: "THE HOLLOW LANTERN"** - Font: Serif with gothic flourishes (e.g., Cinzel, Garamond Gothic, or Bodoni) - Size: Large, occupying upper 15–20% of cover - Color: Warm metallic gold or aged cream with slight transparency - Placement: Top third, centered or slightly off-center - Effect: Subtle shadow or emboss effect for depth **Author: "MARA KVIST"** - Font: Lighter serif or elegant sans-serif (e.g., Lora, Cormorant) - Size: 30–40% smaller than title - Color: Soft cream or pale gray - Placement: Bottom right corner or bottom center - Effect: Minimal; understated elegance --- ## NEGATIVE PROMPT Avoid: watermarks, embedded text within image, bright primary colors, cheerful or whimsical tone, modern settings, cartoon style, cartoon characters, anime, manga, photography, people's faces in sharp focus, copyrighted characters, gore, grotesque imagery, cluttered composition, symmetrical balance, daytime bright sunlight, clear skies, contemporary clothing, overly sharp details, pixelation, low resolution, garish lighting, neon colors. --- ## DIMENSIONS & RESOLUTION **Aspect Ratio:** 3:4 (portrait; standard for print and ebook) **Resolution:** - **Print:** 3000 × 4000 pixels @ 300 DPI (for hardcover/paperback) - **Ebook/Digital:** 1500 × 2000 pixels @ 72 DPI **Safe Zone:** Keep critical image elements within central 85% to avoid cover bleed and spine placement on print editions. --- ## USAGE NOTES This prompt is optimized for **Midjourney v5+** or **Stable Diffusion XL**. Test multiple generations and iterate by adjusting descriptors like "more fog," "stronger lighthouse beam," or "ghostlier mist." Once selected, the image will be layered with typography in design software (Canva, Adobe InDesign, or Affinity Publisher) for final polish.
**Main Prompt** A windswept 19th-century coastal village at dusk beneath storm-heavy skies; a tall, weathered lighthouse on a black cliff emits an eerie hollow lantern glow. In the foreground, a grieving young woman in a dark Victorian cloak stands near wet rocks, back half-turned, hair and hem caught by sea wind, facing the surf as faint ghostlike whispers rise from mist and foam. Crashing waves, slick cobblestones, gulls barely visible, salt haze, distant crooked cottages, and curling sea fog create depth. Muted blue-gray palette with a single warm amber lantern light, cinematic contrast, subtle supernatural atmosphere, elegant negative space for title, highly detailed, painterly realism. **Style & Mood** Gothic mystery, dark fantasy, literary prestige, painterly realism, atmospheric, haunting, melancholic, cinematic. Palette: muted blues, slate, sea-green gray, charcoal, amber gold. Composition: lighthouse dominant upper third, woman foreground lower third, mist framing center. Lighting: storm dusk, volumetric fog, rim light, warm lantern glow. **Typography** Title: large elegant serif, slightly distressed, centered upper-middle or lower-middle; luminous pale ivory with subtle amber accent. Author: smaller uppercase serif or refined small caps, centered at bottom. Strong title hierarchy, generous tracking, classic literary feel. **Negative Prompt** No modern clothing, no modern buildings, no extra people, no bright daylight, no neon colors, no cartoon/anime look, no steampunk gadgets, no ships dominating scene, no visible text in image, no watermark, no low detail, no overly sexualized figure, no symmetrical fantasy castle. **Dimensions** Aspect ratio: 1.6:1 (e.g. 1600×2560) for ebook cover. Print-ready: 6×9 in at 300 DPI = 1800×2700 px minimum; add bleed for full wrap if needed.
### Main Prompt A crumbling 19th-century lighthouse clings to jagged coastal cliffs at twilight, its hollow lantern casting a piercing warm golden glow through fractured glass into swirling sea mist. Below, turbulent waves crash against rocks; a grieving young woman in flowing black Victorian mourning gown stands foreground, windswept hair framing her sorrowful profile gazing seaward, ethereal whisper tendrils rising from foam. Moody fog-shrouded atmosphere. (78 words) ### Style & Mood Painterly digital art in gothic fantasy style akin to Silvia Moreno-Garcia and Erin Morgenstern covers; muted indigo blues, slate grays, cold teal seascape contrasted by volumetric warm amber lantern god rays; deeply atmospheric, melancholic mood; rule-of-thirds composition with woman lower left, lighthouse dominant center-right, dynamic negative space above; cinematic rim lighting on figure, high contrast fog depth. ### Typography Title "The Hollow Lantern" in large, elegant gothic serif font (e.g., like Cinzel or Old Standard TT) with subtle distressing and arched curvature, gold-embossed glow effect, positioned top third spanning width. Author "Mara Kvist" in smaller flowing script font (e.g., Great Vibes), subtle silver, bottom third center. Hierarchy: title 72pt dominant, author 24pt; matte texture overlay for vintage feel; composite over image with 20% opacity drop shadow. ### Negative Prompt blurry, low resolution, modern clothing, bright sunlight, vivid colors, smiling faces, crowds, text, logos, watermarks, cartoons, photorealism, flat lighting, overexposed, deformed anatomy, extra limbs, clear skies. ### Dimensions Aspect ratio 2:3 (portrait, e.g., 1600x2400 pixels for ebook); for print, 1800x2700 at 300 DPI (6x9 inch cover standard). Use --ar 2:3 in Midjourney/Stable Diffusion.
What makes these work
-
01Always specify text-free zones
Most AI models will generate decorative elements across the entire image unless you explicitly tell them not to. State in your prompt where the title and author name will go, such as 'leave the top 30% as a clean gradient sky.' This saves you a post-processing step and prevents critical composition elements from overlapping your typography area.
-
02Name the genre, not just the mood
Prompting 'dark and mysterious' produces generic results. Prompting 'grimdark fantasy in the style of Joe Abercrombie cover art' gives the model a tighter visual target. Genre signals like cozy mystery, Scandinavian noir, or romantasy trigger trained associations that align the output with reader expectations faster than mood words alone.
-
03Use aspect ratio intentionally
Standard trade paperback covers are roughly 6x9 inches, which maps to a 2:3 portrait aspect ratio. Always specify portrait orientation and, if the model supports it, set the exact ratio. A square or landscape output is not usable as a book cover without cropping that destroys your composition.
-
04Iterate on lighting before composition
Lighting determines whether a cover reads at thumbnail size, which is where most buyers see it first. If your first output has flat or ambient light, add specific lighting direction to your next prompt: 'dramatic side lighting from the left,' 'golden hour backlight,' or 'single overhead spotlight.' Composition can be fixed in layout tools. Poor lighting cannot.
More example scenarios
Book cover concept for a dark fantasy novel. A lone armored warrior stands on a cliff edge at dusk, silhouetted against a blood-red sky. Below, a ruined city burns. The mood is grim and epic. No text. Leave the upper third and lower fifth of the image flat and dark to allow title and author name overlays. Style: painterly, high contrast.
A cinematic vertical composition with a backlit silhouette dominating the center, rich crimson and charcoal tones, dramatic volumetric lighting from the burning city below. The upper third fades to near-black and the lower band is sufficiently dark for white text overlay. Genre-legible as epic or grimdark fantasy.
Book cover for a cozy mystery novel set in a fictional English village. A quaint cottage at golden hour with a dead flower arrangement visible through the window. A tabby cat sits on the garden wall looking suspicious. Soft watercolor illustration style, warm palette, slightly whimsical. No text. Leave top 25% open sky for title placement.
A soft-edged watercolor scene with warm amber and sage tones. The cottage is charming but the wilting bouquet in the window adds an unsettling detail. The cat's posture reads as knowing. Sky area is clean and gradient, suitable for dark title text. Clearly signals the cozy mystery genre at thumbnail size.
Book cover for a nonfiction self-help book about deep focus and remote work productivity. Minimal, modern design. A single clean desk with a steaming coffee cup, open laptop glowing, early morning light through a large window. Flat graphic design style, muted blue and white palette, lots of negative space. No text. Keep the top third very clean for large title text.
A minimalist flat-style illustration with cool morning light, a centered desk composition, and significant negative space in the upper half. Blue-white palette reads as calm and productive. The simplicity and negative space make it immediately usable as a nonfiction self-help cover with title text dropped in above the desk scene.
Book cover for a contemporary romance novel set in Bali. Two figures, faces not visible, walking along a rice terrace path at sunset, hands almost touching. Lush green terraces, golden light, distant volcano silhouette. Photorealistic style, warm and dreamy color grading. No text. Composition should allow bottom third for author name in a light script font.
A photorealistic warm-toned image with layered green terraces receding into distance, golden backlight creating a hazy romantic glow. The two figures are intentionally anonymous, which avoids model-face distortion artifacts. Bottom third contains a gradient fade that will accept light-colored script text cleanly. Strong romance genre signal.
Book cover for a near-future sci-fi thriller. A fragmented humanoid face composed entirely of circuit board patterns and glitching pixels, half dissolving into static. Dark background, electric blue and white accent colors, high contrast. Cyberpunk aesthetic, not cartoonish. No text. Upper and lower edges should have dark gradients for title and author overlay.
A high-contrast digital composition with a dissolving face rendered in circuit and glitch textures, electric blue highlights against near-black. The fragmented quality reads immediately as AI or tech-gone-wrong subject matter. Edge gradients fade cleanly to black, leaving usable zones for white or neon-colored title typography. Strong genre signal for tech thriller and cyberpunk readers.
Common mistakes to avoid
-
Asking the model to render readable text
Requesting that the AI include the book title or author name in the generated image almost always produces garbled, misspelled, or distorted lettering. Current diffusion models do not reliably render precise typographic text. Always generate the image text-free and add typography in Canva, Photoshop, or your KDP cover creator.
-
Using a square output for a portrait cover
Defaulting to a square 1:1 image and then cropping it to portrait destroys the original composition and often cuts the focal subject. Set portrait aspect ratio, ideally 2:3, before you generate. Fixing this after the fact almost always requires regenerating from scratch.
-
Prompting mood without specifying genre conventions
A prompt that describes atmosphere without referencing genre conventions produces covers that could belong to any category. Readers make split-second genre identification decisions. If your thriller cover could pass as a children's book because the style is too soft, your prompt lacked specificity about visual conventions for the genre.
-
Treating the first output as final
A single generation is a starting point, not a deliverable. Run at least four to six variations with small prompt adjustments before selecting a direction. Models introduce randomness in each generation, and the third or fourth output often solves a composition problem that the first two had. Using only one output leaves significant quality on the table.
-
Ignoring thumbnail legibility
A cover that looks detailed and impressive at full size often becomes a muddy blur at the 160x250 pixel thumbnail size Amazon and other retailers display. Before finalizing any AI-generated cover concept, shrink it down to thumbnail size and check whether the core image and mood still read clearly. If it does not, simplify the composition.
Related queries
Frequently asked questions
Can I actually publish an AI-generated book cover on Amazon KDP?
Yes, Amazon KDP allows AI-generated covers as long as you have the rights to use the generated image, which depends on the terms of service of the AI tool you used. Many commercial tools, including Midjourney paid tiers and Adobe Firefly, grant commercial use rights. Always verify the specific tool's licensing before publishing. You are also required to disclose AI-generated content in certain contexts, so check current KDP guidelines.
Which AI model is best for generating book covers from text?
Midjourney currently produces the most aesthetically polished and genre-legible results for book cover imagery, particularly for fiction. Adobe Firefly is a strong choice if you need clean commercial licensing with no training data disputes. DALL-E 3 via ChatGPT handles descriptive prompts well but can over-saturate. Stable Diffusion with trained LoRA models can match specific genre aesthetics closely but requires more technical setup.
How do I add the title and author name to an AI-generated cover?
Export the generated image and bring it into a layout tool. Canva has pre-built book cover templates where you can drop your image in and add text layers. For more control, use Adobe Photoshop, Affinity Publisher, or the free tool Scribus. KDP also has a built-in cover creator that accepts uploaded images. Choose a font that matches your genre conventions and ensure sufficient contrast between text color and background.
What image resolution do I need for a publishable book cover?
Amazon KDP requires a minimum of 300 DPI at the final print dimensions, or at least 1000 pixels on the shortest side for ebook covers, with 2560x1600 recommended. Most AI generators output at 1024x1024 or higher, which is adequate for ebook covers. For print, you may need to upscale using a tool like Topaz Gigapixel or Adobe Lightroom's AI upscale to reach 300 DPI at full cover dimensions without visible pixelation.
Do I need design skills to use an AI book cover generator?
Not to generate the base image. Writing a good prompt is a learnable skill that takes practice but does not require formal design training. Where basic design knowledge helps is in the typography step, specifically choosing fonts, sizing, and placement that match genre conventions. If you are unfamiliar with this, use a pre-built Canva book cover template as a starting point and swap in your AI-generated image.
How do I make sure my AI book cover matches my genre's visual conventions?
Study the top 20 bestsellers in your genre on Amazon and identify the recurring visual patterns: color palettes, typography styles, image subject matter, and overall mood. Then reference those conventions directly in your prompt. You can also reference specific cover artists by name in your prompt, such as 'in the style of Drew Struzan' for epic fantasy or thriller, which gives the model a precise stylistic target.