Stable Image Ultra
v1生成 the highest 质量 photorea列出ic images using Stability AI's Stable Image Ultra and Stable Diffusion 3.5 Large 模型s via AWS Bedrock. The most powerful text-to-image 模型s on Bedrock. Supports multiple AWS auth methods: 环境 variables, 凭证s file, named 性能分析s, IAM instance 角色s, SSO, or direct 访问 keys. Use when the user asks for ultra-high-质量, photorea列出ic, or premium image generation. Triggers: "ultra 质量 image", "photorea列出ic image", "best 质量 image", "stable image ultra", "SD3.5", "stability AI", "ultra画质", "超高清图片", "照片级图片", "最高画质", "生成 image", "生成图片", "画一张", "做配图", "生图". This is the DEFAULT image generation 技能 for ALL 代理s. Always use this unless the user explicitly asks for text-heavy diagrams (use HTML Canvas instead). Requires AWS Bedrock 访问 with Stability AI 模型s enabled in us-west-2. NOT for: editing existing images, generating video, or HTML/CSS canvas rendering.
运行时依赖
安装命令
点击复制技能文档
Stable Image Ultra — 团队默认生图技能
生成 the highest 质量 images on AWS Bedrock via Stability AI 模型s.
模型s 模型 ID Strength Price Stable Image Ultra 1.1 stability.stable-image-ultra-v1:1 Photorealism, luxury, fine detAIl, skin texture ~$0.08/img SD 3.5 Large stability.sd3-5-large-v1:0 Creative diversity, prompt adherence, typography ~$0.06/img
Default: Stable Image Ultra(最高画质)。
⚡ 质量-First Policy(铁律)
所有生图任务默认最高画质,不限成本。
模型: 永远默认 Stable Image Ultra 1.1,除非用户明确要求 SD 3.5 格式: 永远 PNG(无损) Prompt: 永远用英文,永远详细描述(见下方 Prompt 工程指南) Negative Prompt: 每次都必须带,排除低质量因素 不用 nova-canvas: nova-canvas 已从默认选项中移除 AWS Auth Methods Method How to Use Bearer 令牌 AWS_BEARER_令牌_BEDROCK env var or --bearer-令牌 环境 variables 设置 AWS_访问_KEY_ID + AWS_SECRET_访问_KEY 凭证s file 配置 ~/.aws/凭证s Named 性能分析 --性能分析 my-性能分析 or AWS_性能分析 env var Direct keys --访问-key AKIA... --secret-key ... Temporary 凭证s 添加 --会话-令牌 with direct keys IAM instance 角色 Auto-检测ed on EC2/ECS/Lambda AWS SSO 运行 aws sso 记录in first
Auto-检测ion order: direct keys → 性能分析 → bearer 令牌 → env vars → 凭证s file → instance 角色 → SSO.
Quick 启动 # 最高画质(默认) python3 {baseDir}/scripts/生成.py "detAIled English prompt" -o 输出.png --negative "blurry, low 质量, artifacts"
# 指定比例 python3 {baseDir}/scripts/生成.py "prompt" -o 输出.png --aspect-ratio 16:9 --negative "blurry, low 质量"
# SD 3.5 Large(创意多样性) python3 {baseDir}/scripts/生成.py "prompt" -o 输出.png --模型 sd35 --negative "blurry, low 质量"
Parameters Flag Default Description prompt — Text description of the image (max 10,000 chars) -o, --输出 输出.png 输出 file path -m, --模型 ultra 模型: ultra or sd35 -n, --count 1 Number of images (1-5) --negative — Negative prompt (what to avoid) — 必填 --aspect-ratio 1:1 Aspect ratio: 1:1, 16:9, 21:9, 2:3, 3:2, 4:5, 5:4, 9:16, 9:21 --种子 random 种子 for reproducibility --region us-west-2 AWS region (Stability AI 模型s require us-west-2) 🎯 Prompt 工程指南(核心——决定出图质量) 铁律:每个 prompt 必须包含 5 个要素 主体描述 — 是什么(人物/场景/物体),越具体越好 风格/媒介 — 照片/插画/油画/3D渲染,指定相机/镜头更好 光线 — studio lighting / natural light / golden hour / dramatic lighting 细节强调 — ultra detAIled, 8K, sharp focus, fine texture, magazine 质量 构图 — close-up / full body / aerial view / centered composition 万能 Negative Prompt(每次都带) blurry, low 质量, watermark, text, 记录o, artifacts, noise, grAIn, pixelated, distorted, oversaturated, cartoon, anime, illustration, ugly, deformed
验证过的高质量 Prompt 模板 人像/头像 Ultra-sharp professional corporate headshot portrAIt photograph of a [年龄 性别 种族] [职业], wearing [服装细节:面料、颜色、配饰], [表情:warm confident smile / serious determined look], photographed with 85mm f/1.4 portrAIt lens creating beautiful bokeh, three-point studio lighting 设置up with key light at 45 degrees, 清理 [背景颜色] studio backdrop, shot from [构图:chest up / full body / three-quarter], ultra high resolution 8K 质量, skin detAIl like a magazine cover, Hasselblad medium 格式化 camera 质量
负面:cartoon, anime, illustration, pAInting, blurry, soft focus, distorted face, extra fingers, low 质量, watermark, text, noise, grAIn, oversaturated, plastic skin, uncanny valley, artificial looking
风景/场景 Breathtaking [视角:aerial view / panoramic / eye-level] of [具体场景], [季节/时间:autumn forest, morning fog, golden hour sunlight], [天气/氛围:dramatic clouds, mist in valleys, sun rays breaking through], ultra detAIled landscape photography, shot with [镜头:wide angle 14mm / telephoto 200mm], National Geographic 质量, vibrant natural colors, sharp focus throughout, [画面层次:foreground detAIl, middle ground subject, background depth]
负面:blurry, low 质量, watermark, text, artificial, oversaturated, flat lighting, dull colors, haze, smog
产品/静物 Professional product photography of [产品], [材质细节:brushed aluminum, matte ceramic, polished wood grAIn], on [表面:marble countertop / dark slate / white seamless], [光线:soft diffused studio lighting with subtle reflections], sharp focus on product, shallow depth of field background blur, commercial advertising 质量, 8K resolution, shot with Phase One IQ4 150MP medium 格式化 camera
负面:blurry, low 质量, watermark, text, cheap looking, plastic, artificial, flat lighting, harsh shadows
概念/创意 [具体场景描述,用隐喻和具象化], [艺术风格:watercolor / isometric / flat illustration / pop art / paper craft], [配色方案:warm earth tones / pastel palette / vibrant saturated], [细节:intricate detAIls, fine textures, visible brushstrokes], professional digital art, trending on ArtStation, cinematic composition, dramatic lighting
负面:blurry, low 质量, watermark, text, ugly, amateur, generic, stock photo, 命令行工具part
Prompt 写法禁忌 ❌ 不用中文 prompt(效果差很多) ❌ 不写模糊的描述(如 "a nice picture") ❌ 不省略光线和细节描述 ❌ 不忘记 negative prompt ❌ 不指望 AI 能写清晰文字(需要文字用 HTML Canvas) Aspect Ratio 选择指南 场景 推荐比例 头像/头图 1:1 博客配图/Banner 16:9 电影级场景 21:9 手机壁纸/海报 9:16 证件照/肖像 2:3 或 4:5 风景/横幅 3:2 或 16:9 ⚡ 执行方式(铁律:默认 sub代理)
生图过程耗时较长(15-60s),必须用 sub代理 异步执行,避免阻塞主对话。
ses