
Google Imagen4
从文字描述开始生成图片,默认采用 1:1 方形比例。可按用途选择横向、竖向或自适应比例,也可填写反向提示词和文本随机种子。
快速了解
- 主提示词最多 5,000 个字符,无需准备参考图。
- 默认 1:1,可选 16:9、9:16、3:4、4:3 或自适应。
- 每次任务预计 8 积分,已开放的比例、反向提示词和随机种子选项不改变预计金额。
相关查询: 如何在 Kyeo 使用 Google Imagen4 · Google Imagen4 每次多少积分 · Google Imagen4 提示词上限是多少
示例


Fantasy scene

Visual concept
Ultra-high-fashion editorial beauty portrait of a fair-skinned female model with sleek center-parted hair tightly pulled into a low bun, wearing a fitted black ribbed turtleneck sweater. Dramatic studio lighting creates a bold diagonal shadow slicing across the face, emphasizing facial structure, freckles, and glossy deep-red lips. Minimal makeup with natural skin texture, cinematic contrast, rich warm color palette. Background consists of striking geometric color-blocked panels in vibrant orange and deep burgundy red, forming strong diagonal lines. Clean composition, luxury magazine cover aesthetic, contemporary fashion campaign, fine-art portrait photography, sharp focus, ultra-detailed skin, professional studio setup, high dynamic range, moody atmosphere, Vogue-style editorial, medium close-up framing, 85mm lens, f/2.8, photorealistic, 8K quality. Negative Prompt: blurry, low resolution, overexposed highlights, flat lighting, messy hair, smiling expression, heavy accessories, jewelry, cluttered background, text, watermark, logo, extra limbs, distorted face, cartoon, illustration, CGI, oversaturated skin, unnatural anatomy, poor composition, noise, motion blur.

Fantasy scene

Visual concept
Ultra-realistic design culture magazine cover featuring a creative Korean woman in a bold yellow structured blazer and black wide-leg trousers — matching the product's colorway — holding a Teenage Engineering OP-1 Field synthesizer horizontally with both hands, fingers hovering over the tiny keys mid-composition, deep in creative flow. Bold graphic makeup, sharp geometric liner, yellow eyeshadow accent. Sleek straight bob. Background: clean white Bauhaus-inspired studio with bold yellow color block wall panels, ultra-graphic composition. "DEZEEN" masthead in precise sans-serif with "INSTRUMENTS OF DESIGN ISSUE" callout. 50mm lens, ultra-detailed OP-1 tiny key and OLED display, 8K photorealistic design editorial. Negative: cartoon, blurry, incorrect product shape, watermark, distorted.

Visual concept
Use the provided image strictly as the reference for the subject’s face, age, and identity. Lock facial identity with the highest priority. Preserve the exact facial structure, proportions, skin tone, jawline, eyes, nose, lips, and overall likeness. Do not beautify, stylise, feminise beyond the natural identity, or alter facial features in any way. Create a modern, high-end studio portrait of a woman in a vertical 4:5 aspect ratio. Full-body seated editorial composition, centre-framed with balanced negative space. The overall visual aesthetic should feel clean, luxurious, powerful, and magazine-quality. The subject is wearing a sharply tailored deep forest green women’s suit with a refined structured blazer and tailored straight-leg trousers, paired with a crisp black fitted button-down shirt, elegant black leather formal heels, and a black wristwatch. The overall color palette is rich deep forest green (#225d57) with a polished luxury editorial aesthetic. The subject is seated in a sleek modern white molded plastic chair with four thin metal legs. Her posture is relaxed and confident, leaning back comfortably with her back supported by the chair rather than sitting rigidly upright. Torso angled slightly toward the viewer’s left side. Head turned left, showing a refined three-quarter profile. Gaze directed toward the left side of the frame, not toward the camera, creating a thoughtful editorial look. Body language: Shoulders relaxed and naturally dropped. Right hand resting near the upper thigh, fingers relaxed and naturally extended. Left elbow resting lightly on the edge of the seat, with the wristwatch visible. Legs extended forward in a relaxed seated stance. Both feet resting naturally on the floor, slightly apart and angled toward the left side of the frame. The pose should feel elegant, grounded, natural, and editorial — never stiff. The chair is placed in a seamless dark green studio environment where the floor and background blend smoothly. Behind the subject’s head and upper torso is a medium-sized glowing white circular LED ring light, centered like a modern halo. This creates a strong graphic visual element and subtle rim-light separation from the background. Use cinematic studio lighting: soft key light from the left side shaping the face, gentle shadow retention for depth, and subtle controlled fill light to preserve detail. The LED ring behind the subject should glow clean and bright without overpowering the subject. The lighting mood should feel dramatic, premium, and polished. Photorealistic, ultra-detailed, premium commercial portrait, sharp focus on facial features, suit texture, shoes, chair surface, and lighting contrast, refined editorial color grading, clean studio finish, no bad anatomy or bad proportions, no distortion, no exaggerated limbs or hands.
一句话了解
这个模型怎么样?
先写清画面里要出现什么,再选择比例。反向提示词用于描述希望避免的内容,最多 5,000 个字符;随机种子在这个入口是可选文本,最多 500 个字符。
- 模型定位
- 工作台只接受文字描述,可选择比例并填写反向提示词、文本随机种子。这个入口没有图片上传、局部编辑、输出张数、分辨率、文件格式、安全设置或自动改写提示词选项。
- 适合谁用
- 想先做方形画面时,可以从默认 1:1 开始。需要其它版式,再从 5 种固定比例或自适应中选择,检查主体位置、裁切、文字和留白是否合适。
- 本站使用说明
- 提交前可以查看预计积分,生成后在任务提示和积分明细中核对预扣、确认消耗及退还记录。结果已生成但消耗尚未确认时,仍会保留预扣积分并提示你到明细中核对。
关键参数一览
根据这些信息,判断模型是否适合你的任务。
它还叫什么
同一个模型在公开资料中可能有不同名称,这里整理相关称呼,方便查找和比较。
用户常见问题
从任务、输入条件和结果要求出发,整理选择模型时需要确认的问题。
选型指南
根据输入素材、任务要求和积分成本,逐项确认模型是否适合。
没有参考图、希望从文字描述开始时,可以先保留默认 1:1。提交前确认模型可用,并查看当前预计的 8 积分;不要仅凭型号名称判断画面效果。
这里没有图片上传和局部编辑选项。若要以原图为基础修改、放大图片或指定输出格式,请先检查其它模型页面是否提供所需操作及输入条件。
可用最多 5,000 个字符描述希望避免的内容,也可填写最多 500 个字符的文本随机种子。先记录输入,再观察结果;这些选项不保证排除指定内容或重现同一张图片。
模型对比
把常见候选放在同一任务下比较,帮助你确认输入方式、控制项和成本差异。
模型横向对比表
按同一组维度比较当前模型和候选模型的输入、设置与使用边界。
适用情况
依据公开资料与当前站内设置,说明模型适用的任务及需要注意的限制。
先查看当前可用状态
比较实际选项与预计积分
能力详解
用文字描述画面
宽高比
随机种子
适合做什么
以 1:1 为起点的比例比较
反向提示词与 seed 对照
与 Fast、Ultra 比较可选设置
提示词技巧
主提示词先写可见内容
反向提示词写要避免的内容
比较随机种子时记录其它设置
为什么选它
使用前要知道
可调参数
根据这些信息,判断模型是否适合你的任务。
积分消耗
每次任务预计 8 积分,任务受理后预扣。选择比例、填写反向提示词或文本随机种子不改变这笔预计积分;实际消耗确认后按确认金额结算。前置校验或限流未通过时,不会预扣生成积分。
准备多次生成时,建议先用少量素材测试,核对输入设置和结果是否适合任务,再决定是否继续。每次生成均按工作台显示的计费规则扣点。
常见问题
相关模型
按你的任务比较这些模型的输入要求、功能和当前状态,再查看对应详情。
Google Imagen4 Fast
用提示词生成图片,画幅可选 1:1、16:9、9:16、3:4、4:3 或自适应,默认 16:9。本页不接收图片,也不提供图片编辑。
Google Imagen4 Ultra
Google Imagen4 Ultra 根据文字描述生成图片,默认画幅为 1:1。你可以选择其它画幅,填写希望避免的内容或文本随机种子。任务受理后先预扣 12 积分。
信息来源
页面内容依据公开模型资料与 Kyeo AI 的模型信息整理。能否在线使用,请以当前模型状态为准。
本页参考 Google Cloud 的 Imagen 4 型号说明、文生图指南,以及 Google 官方服务更新资料。官方内容用于介绍型号背景和服务范围;Kyeo AI 的输入选项、积分规则及可用状态仍需按本站当前页面与工作台核对,不能由官方停用或迁移安排直接推断。
说明 Google Cloud 的 Imagen 4 型号与服务范围;本站输入选项、计费和可用状态仍以当前页面为准。
Vertex AI 在 2026 年 3 月 24 日公告中宣布 Imagen 4 系列相关型号弃用,建议在 6 月 30 日前迁移;这不是 Kyeo AI 的停用日期。
介绍 Google Cloud 的文字生成图片用法,不代表本站提供文档中的全部参数。
2026 年 6 月 15 日公告将 Imagen 4 系列在 Gemini API 的计划停用日期列为 8 月 17 日;该公告不说明 Kyeo AI 的停用安排。
当前说明 Imagen 已不在 Gemini API 提供,需与本站模型页面的可用状态分别查看。