[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC)：2025-08-25。"],[],[],null,["# Migrate to Imagen 3\n\nThis guide explains how to migrate your applications from\nImagen 1 and Imagen 2 to\nImagen 3.\n\nWhy migrate to Imagen 3?\n------------------------\n\nStarting on June 24, 2025, Imagen versions 1 and 2, Image\nCaptioning, and Visual Questioning Answering are deprecated.\nImagen models `imagegeneration@002`, `imagegeneration@005`,\nand `imagegeneration@006` will be removed on September 24, 2025 .\n\nImagen 3 offers performance improvements over\nImagen 1 and Imagen 2.\n\nUpgrading most Imagen applications to\nImagen 3 shouldn't require significant reengineering of\nprompts or code. But some applications may require prompt changes, and these\nchanges are difficult to predict without running a prompt through\nImagen 3 first.\n\nWhich Imagen 3 model should I migrate to?\n-----------------------------------------\n\nThe Imagen workloads that you use determine which\nImagen 3 model we recommend for your migration.\n\nThe following table shows common workloads for Imagen 1 and\nImagen 2 models and their support by\nImagen 3 models:\n\nHow do I migrate?\n-----------------\n\nIn most cases you can update the model name that you are using to use an\nImagen 3 model with no further changes to your parameters.\nThe following sections provide exceptions that may affect you.\n\n### Image Captioning and Visual Question Answering\n\nIf you use Imagen's [Image Captioning and Visual Question\nAnswering](/vertex-ai/generative-ai/docs/image/image-captioning)\nfeatures, we recommend that you migrate to Gemini Image Understanding\nfor improved capabilities for analyzing and describing images.\n\nFor more information, see [Image\nunderstanding](/vertex-ai/generative-ai/docs/multimodal/image-understanding).\n\n### Inpainting, outpainting, and product editing\n\nInpainting, outpainting, and product editing require API changes when migrating\nto Imagen 3 models. For more information, see the following:\n\n- [Inpainting insert content (bring your own\n mask)](/vertex-ai/generative-ai/docs/image/edit-inpainting#inpaint-insert-mask-imagen3-tab)\n\n- [Outpainting](/vertex-ai/generative-ai/docs/image/edit-outpainting#outpainting-imagen3-tab)\n\n- [Product image editing (automatic background\n detection)](/vertex-ai/generative-ai/docs/image/edit-product-image#product-image-editing-imagen3-tab)\n\n### Personalization\n\nFor avatar stylization, migrate from [Imagen 2\nPersonalization](/vertex-ai/generative-ai/docs/image/edit-personalization) to\nImagen 3's [Subject\ncustomization](/vertex-ai/generative-ai/docs/image/subject-customization), which\nprovides improved capabilities.\n\nAre there pricing changes?\n--------------------------\n\nThere may be pricing changes for upgrading from earlier Imagen\nmodels.\n\nFor information about pricing, see the following:\n\n- [Imagen](/vertex-ai/generative-ai/pricing#imagen-models)\n- [Veo](/vertex-ai/generative-ai/pricing#veo)\n- [Gemini](/vertex-ai/generative-ai/pricing#gemini-models)"]]
