返回所有教程
2026-01-12
如何从文本生成AI图像
文本到图像的关键是提示词清晰度。使用简短的提示词:主体 + 构图 + 光照 + 风格,然后每次迭代一个更改。
快速步骤
一分钟内可以完成的快速清单。
- 1打开文本到图像工具。
- 2编写简短提示词:主体 + 场景 + 光照 + 风格。
- 3如果可用,添加负面提示词(以减少瑕疵)。
- 4生成2-4个变体。
- 5通过每次更改一个细节(相机、光照、风格)来迭代。
Tool options you can choose
Aspect ratio:1:12:33:24:33:44:55:49:1616:9
- Output format: PNG.
教程示例(包含提示词和设置)
以下每个示例都是为本指南预先选择的(非随机)。
Example 1
Photoreal example prompt
Text-to-Imagetext-to-image
How to use this example
- 1.Open the tool.
- 2.Choose mode: Text-to-Image.
- 3.No input file needed (prompt-only).
- 4.Use the keywords (or full prompt) and pick settings.
- 5.Generate and iterate (crop/lighting/prompt) if needed.
Prompt keywords
Raindrops fallcreating concentric ripplesWater splashes on the foreground edge of the lensforming realisticwet halos and tiny dropletsA person in a dark trench coat walks past with their back to the camerathe edge of the coat is lifted by the windrevealing the fabric's texture clearly
Show full prompt (copy)
Raindrops fall, creating concentric ripples, and the red and blue reflections of neon signs are slightly distorted on the water's surface. Water splashes on the foreground edge of the lens, forming realistic, wet halos and tiny droplets. A person in a dark trench coat walks past with their back to the camera; the edge of the coat is lifted by the wind, revealing the fabric's texture clearly. Their footsteps step into puddles, and the splashing water droplets shimmer like tiny silver fragments in the backlight of the streetlights. The background is a slightly bokehd neon light and mist, with fine rain and a thin fog floating in the air, and volumetric light passing through the alleyway to form visible beams of light.
Settings (used in this example)
Notes
Negative: lowres, blurry, overexposed, oversaturated, messy clutter, cheap plastic look, cartoon, anime, CGI, distorted perspective, warped geometry, extra objects, text, logo, watermark, harsh flash, unnatural skin/lighting.
Example 2
Cinematic interior prompt
Text-to-Imagetext-to-image
How to use this example
- 1.Open the tool.
- 2.Choose mode: Text-to-Image.
- 3.No input file needed (prompt-only).
- 4.Use the keywords (or full prompt) and pick settings.
- 5.Generate and iterate (crop/lighting/prompt) if needed.
Prompt keywords
Cinematic high-end editorial sceneultra-premium aesthetic: a modern luxury penthouse lounge at golden hourfloor-to-ceiling windows overlooking a soft-focus city skylinewarm sunbeams cutting through light hazenatural rim lightsubtle film grainMinimalist interior with travertine stonebrushed brass accents
Show full prompt (copy)
Cinematic high-end editorial scene, ultra-premium aesthetic: a modern luxury penthouse lounge at golden hour, floor-to-ceiling windows overlooking a soft-focus city skyline, warm sunbeams cutting through light haze, natural rim light, subtle film grain. Minimalist interior with travertine stone, brushed brass accents, dark walnut wood, cream boucle sofa, a single sculptural vase with white calla lilies, curated art book on a marble coffee table. Clean composition, negative space, calm mood, tasteful luxury, no clutter. Photorealistic, high dynamic range, crisp details, soft shadows, 85mm lens look, shallow depth of field, cinematic color grading, glossy magazine cover quality, 8k.
Settings (used in this example)
Notes
Negative: lowres, blurry, overexposed, oversaturated, messy clutter, cheap plastic look, cartoon, anime, CGI, distorted perspective, warped geometry, extra objects, text, logo, watermark, harsh flash, unnatural skin/lighting.
提示
- 简短的提示词通常在清晰度上优于长提示词。
- 指定相机术语(特写、85mm、浅景深)以增加真实感。
- 如果输出混乱,移除多余物体并简化场景。
常见问题
提示词应该用英语吗?▼
通常是的——英语提示词在各种模型中往往更可靠。
什么提示词结构效果最好?▼
主体 + 动作/姿势 + 相机 + 光照 + 风格。保持简洁。
如何减少图像中的标志/文字?▼
Use negative prompts like “text, logo, watermark” and keep backgrounds simple.
准备好生成了吗?
打开工具并重用上面的提示词/设置。