
面向开发者的 AI API
图生图 API
通过 PoYo 图生图 API 接入 AI 图片编辑。以产品照片、人物肖像或设计草图为参考,对比编辑控制、主体保持、输出质量与价格。
匹配模型
32 个模型

$0.04 /次生成
nano-banana-pro
Google's advanced Gemini 3 Pro image model featuring 4K resolution, enhanced reasoning, real-time data integration, and multi-image composition capabilities.

$0.01 /次生成
gpt-image-2
OpenAI's premium image generation family for text-to-image creation and single-image editing with simple prompt-driven controls.

$0.024 /张
qwen-image-3
Alibaba's third-generation image model for long prompts, information-dense layouts, fine text rendering, multilingual visuals, and knowledge-rich image generation.

$0.009 /张
flux-3-klein
Generate and edit images with FLUX 3 Image Klein. Use 1K, 2K or 4K output, up to ten editing references, custom dimensions, and PNG, JPEG or WebP output.

$0.075 /张
seedream-5.0-pro
ByteDance's flagship image model for precise prompt following, dense layouts, multilingual text rendering, and region-precise editing.

$4.00 100 万输入
gpt-image-2.5-flare
OpenAI image generation and editing with high-fidelity reference images, precise inpainting, multi-turn consistency, and transparent backgrounds. Explore Flare and Sunburst and compare GPT Image 2.5 with GPT Image 2 on PoYo.

$4.00 100 万输入
gpt-image-2.6-flare
Affordable GPT Image 2.6 API for text-to-image and reference image editing with OpenAI's Flare and Sunburst. One REST call, from 1.7 credits ($0.0085) per image on PoYo.

$0.04 /张
grok-imagine-image-2.0
xAI image generation and editing with up to three input images, five supported aspect ratios, 1K or 2K resolution, selectable quality, and up to four outputs.

$1.40 100 万输入
mai-image-2.6
MAI Image 2.6 and MAI Image 2.6 Flash API on PoYo: Microsoft's image models for photorealistic text-to-image, legible in-image text, and editing or composition with up to 5 reference images, with optional web search grounding.

$0.016 /张
minimax-h3-image
MiniMax H3 Image API on PoYo: MiniMax's open-weight image model for text-to-image and editing with up to 16 reference images, 1K, 2K or 4K output, custom sizes up to 20 megapixels, and sharp in-image text. From 3.2 credits ($0.016) per image.

$0.03 /次生成
qwen-image-2.1
Alibaba's 7B open-weight unified text-to-image and editing model with native RGBA transparency, up to 10 reference images, and 2K resolution.

$0.008 /张
seedream-5.5
Seedream 5.5 API on PoYo: ByteDance Seed's image model for text-to-image, multi-reference editing with up to 10 images, native transparent PNGs, and layer decomposition into a base image plus up to 99 layers. From 4 credits ($0.02) per image.

即将上线
MAI Image 2.7
Microsoft's editing-focused MAI Image 2.7 preview model. It has no public API yet; coming soon to PoYo. MAI Image 2.6 is available now.

$0.008 /张
ideogram-v4.5
Generate images, posters and typography with Ideogram 4.5, or precisely edit a source image with masks and reference images. Compare quality-based pricing for generation and editing on PoYo.

$0.10 /张
ideogram-vector-1.0/text-to-vector
Ideogram Vector creates editable SVG artwork from a text prompt, with optional reference images, or converts one image into an SVG. Text to vector is billed per SVG at the same rates Fal publishes for low, medium, and high quality.

$0.02 /张
nano-banana-2.1
Generate and edit 1K, 2K and 4K images with Google's Nano Banana 2.1. Combine up to 10 reference images for product visuals, posters and consistent creative assets. From 4 credits ($0.02) per image.

$0.02 /张
seedream-5.0-flash
ByteDance's fast, cost-efficient image model for 1K/1.5K/2K generation, multi-reference editing, and multilingual visual content. 4 credits per image.

$0.03 /次生成
flux-2-pro
Black Forest Labs' production-grade model combining 4MP image generation and editing with multi-reference support, precise typography, and hex color control.

$0.01 /次生成
gpt-image-1.5
OpenAI's latest image model with 4x speed, precision editing, and superior text rendering.

$0.025 /次生成
nano-banana
Google's leaderboard-topping image model (Gemini 2.5 Flash) excelling in natural language editing, character consistency, and multi-image blending.

$0.025 /次生成
seedream-4
ByteDance's multimodal image model for native 1K to 4K generation, reference-guided editing, and batch outputs up to 15 images.

$0.04 /次生成
flux-kontext-pro
Black Forest Labs' in-context image generation and editing family for scene-preserving edits, text replacement, and prompt-driven visual changes.

$0.02 /次生成
gpt-4o-image
OpenAI's native multimodal image generator with exceptional text rendering, precise prompt following, and conversational editing capabilities.

$0.025 /次生成
seedream-4.5
ByteDance's unified 4K image generation and editing model with professional-grade text rendering and commercial photography quality.

$0.025 /次生成
nano-banana-2-new
Google's next-gen image model powered by Gemini 3.1 Flash with native 2K/4K resolution, chain-of-thought reasoning, precise multi-language text rendering, and up to 14 reference images.

$0.025 /次生成
seedream-5.0-lite
ByteDance's efficient multimodal image model with native 2K/3K resolution, deep thinking, bilingual text rendering, and smart editing.

$0.021 /次生成
wan-2.7-image
Alibaba's unified Wan 2.7 image family for text-to-image generation and reference-based image editing with standard and pro quality tiers.

$0.025 /次生成
nano-banana-2-lite
Google's fastest and most cost-efficient Gemini Image model, built for rapid ideation, high-throughput image generation, low-latency creative workflows, and scaled production pipelines.

$0.018 /张
kling-o3-image
Kling's O3 image family supports prompt-only generation, multi-reference editing, single outputs, and connected series generation with 1K to 4K resolution tiers.

$0.018 /张
kling-o1-image-edit
Kling O1 Image Edit focuses on reference-based image transformation with optional elements guidance, flexible aspect ratios, and 1K or 2K output tiers.

$0.02 /百万像素
flux-dev
Black Forest Labs image model for text-to-image generation and single-image editing with output-size based pricing.

$0.04 /张
grok-imagine-image-quality
xAI's higher-fidelity Grok Imagine image model for polished text-to-image generation and reference-based image editing.
图像生成图像模型 API - 价格和模型适配
选择模型前,对比供应商、支持的输入、输出格式和价格。
精确任务匹配
只有模型目录元数据包含 Image to Image 的模型才会出现在这里。
供应商对比
无需离开任务目录,即可查看不同供应商的模型系列。
API 就绪路径
每张模型卡都会跳转到 PoYo 的价格、示例、Playground 控件和 API 详情。
| 模型 | 供应商 | 任务类型 | 价格 |
|---|---|---|---|
| Nano Banana Pro nano-banana-pro | Text to Image, Image to Image | $0.04/次生成 | |
| GPT Image 2 gpt-image-2 | OpenAI | Text to Image, Image to Image | $0.01/次生成 |
| Qwen Image 3.0 qwen-image-3 | Alibaba | Text to Image, Image to Image, Uncensored | $0.024/张 |
| FLUX 3 Image Klein flux-3-klein | Black Forest Labs | Text to Image, Image to Image | $0.009/张 |
| Seedream 5.0 Pro seedream-5.0-pro | Seedream | Text to Image, Image to Image, Uncensored | $0.075/张 |
| GPT Image 2.5 gpt-image-2.5-flare | OpenAI | Text to Image, Image to Image | $4.00100 万输入 |
| GPT Image 2.6 gpt-image-2.6-flare | OpenAI | Text to Image, Image to Image | $4.00100 万输入 |
| Grok Imagine Image 2.0 grok-imagine-image-2.0 | xAI | Text to Image, Image to Image | $0.04/张 |
| MAI Image 2.6 mai-image-2.6 | Microsoft | Text to Image, Image to Image | $1.40100 万输入 |
| MiniMax H3 Image minimax-h3-image | MiniMax | Text to Image, Image to Image | $0.016/张 |
| Qwen Image 2.1 qwen-image-2.1 | Alibaba | Text to Image, Image to Image | $0.03/次生成 |
| Seedream 5.5 seedream-5.5 | Seedream | Text to Image, Image to Image | $0.008/张 |
| MAI Image 2.7 mai-image-2-7 | Microsoft | Image to Image | 即将上线 |
| Ideogram 4.5 ideogram-v4.5 | Ideogram | Text to Image, Image to Image | $0.008/张 |
| Ideogram Vector 1.0 ideogram-vector-1.0/text-to-vector | Ideogram | Text to Image, Image to Image | $0.10/张 |
| Nano Banana 2.1 nano-banana-2.1 | Text to Image, Image to Image | $0.02/张 | |
| Seedream 5.0 Flash seedream-5.0-flash | Seedream | Text to Image, Image to Image | $0.02/张 |
| FLUX.2 flux-2-pro | Black Forest Labs | Text to Image, Image to Image | $0.03/次生成 |
| GPT Image 1.5 gpt-image-1.5 | OpenAI | Text to Image, Image to Image | $0.01/次生成 |
| Nano Banana nano-banana | Text to Image, Image to Image | $0.025/次生成 | |
| Seedream 4 seedream-4 | Seedream | Text to Image, Image to Image | $0.025/次生成 |
| Flux Kontext flux-kontext-pro | Black Forest Labs | Text to Image, Image to Image | $0.04/次生成 |
| GPT-4o Image gpt-4o-image | OpenAI | Text to Image, Image to Image | $0.02/次生成 |
| Seedream 4.5 seedream-4.5 | Seedream | Text to Image, Image to Image, Uncensored | $0.025/次生成 |
| Nano Banana 2 nano-banana-2-new | Text to Image, Image to Image | $0.025/次生成 | |
| Seedream 5.0 Lite seedream-5.0-lite | Seedream | Text to Image, Image to Image, Uncensored | $0.025/次生成 |
| Wan 2.7 Image wan-2.7-image | Alibaba | Text to Image, Image to Image | $0.021/次生成 |
| Nano Banana 2 Lite nano-banana-2-lite | Text to Image, Image to Image | $0.025/次生成 | |
| Kling O3 Image kling-o3-image | Kling | Text to Image, Image to Image | $0.018/张 |
| Kling O1 Image kling-o1-image-edit | Kling | Image to Image | $0.018/张 |
| Flux Dev flux-dev | Black Forest Labs | Text to Image, Image to Image | $0.02/百万像素 |
| Grok Imagine Image Quality grok-imagine-image-quality | xAI | Text to Image, Image to Image | $0.04/张 |
常见问题
什么是图生图 API?
+
图生图 API 以来源图片和编辑指令生成修改结果或视觉变体。图片提供主体、布局和风格信息,具体参考程度及可用编辑模式取决于所选模型。
如何向图片编辑 API 提交图片?
+
使用接口文档列出的图片字段与编码方式。需要图片 URL 时,确认文件无需浏览器登录即可访问,同时检查格式、尺寸和文件大小;私人会话中的链接不一定能被服务端读取。
如何修改背景并保留产品或人物?
+
提供清晰来源图,明确要求保留的人脸、包装、颜色、标志或产品结构。每次先做一项主要修改,并仔细检查输出;保留主体的指令不能保证所有像素和细节完全不变。
所有图生图 API 都支持蒙版和局部重绘吗?
+
不支持。参考生成、指令编辑、蒙版重绘和扩图是不同能力。先确认接口是否提供所需模式和字段,不能通过添加未记录的蒙版参数来假定局部修改会生效。
可以使用多张参考图吗?
+
部分模型支持多图,数量限制和参考方式因接口而异。可按支持情况说明每张图负责产品外观、构图或风格,避免相互冲突的参考,并提前确认图片数量限制。
图生图提示词怎么写更有效?
+
分别说明需要改变什么、保留什么。例如替换背景,同时保留产品形状、颜色和标签。避免一次加入过多无关修改,逐步迭代更容易定位不理想的原因。
图生图 API 可以放大图片吗?
+
部分模型支持更大的输出尺寸,但重新生成也可能改变画面内容。区分图片编辑、新画面生成与忠实放大,分别评估尺寸和细节保持;必要时选择专门的增强工具。
图生图与文生图有什么区别?
+
图生图以已有视觉素材为起点,适合产品修订、人像编辑和参考生成;文生图从文字描述构建新画面。需要维持具体主体或设计时,参考图通常更有帮助。
如何通过 API 获取编辑结果?
+
按模型文档提交图片输入,保存 task_id,通过状态查询或支持的 webhook 获取完成结果。在应用中保存来源图片与输出的对应关系,便于比较版本和追踪修改。
图生图 API 价格受什么影响?
+
模型、输出尺寸、图片数量和可选编辑能力都可能影响费用,部分接口还单独计算参考输入或尺寸处理。查看对应计费档位,并把获得可用结果所需的尝试次数计入成本。
为什么不想修改的细节也变了?
+
模糊参考或相互冲突的指令可能导致更广泛的重新生成。尝试更清晰的图片、缩小修改范围并明确保留内容;问题持续时,可比较专用编辑模式或其他模型。
在哪里查找图片编辑 API 参数和示例?
+
进入所选模型页查看当前图片字段、编辑设置和请求示例,也可读取模型专属 llms.txt。按该接口的结构接入,不要假定所有图片编辑模型接受相同参数。