National Gallery Mixtape 是伦敦国家美术馆联合 Google Arts & Culture 推出的 AI 音乐实验工具。汇集了伦敦国家美术馆精选的200幅世界名画，涵盖从文艺复兴时期到现代的各种风格和主题。用户可从这些画作中最多选择6幅，拖放到指定区域，AI会分析画作的色彩、主题、情感和历史背景等元素，生成与之匹配的音乐片段。能通过调整音乐片段的音量、顺序和叠加方式，以及使用音乐风格、乐器和心情标签等工具来定制音乐的风格和氛围。
官网地址： https://artsandculture.google.com/experiment/iwHPaSXS-zrryg
<p>字节开源DreamO，统一图像定制框架，把图像换装、换脸、换造型、换风格以及组合操作装在了一起 支持ID、IP、Try-On等组合，支持16GB/24GB显卡运行，用于虚拟试穿、商品广告、营销广告什么的比较实用 四个能力： IP，处理角色形象，支持人物、物体、动物等输入 ID，人脸身份处理 Try-On，虚拟试穿，可以同时换多件衣服 Style，风格迁移，目前还不能和其他任务组合</p> <p>DreamO正式实现<a href="https://arxiv.org/abs/2504.16915" rel="nofollow">：统一的图像定制框架</a> </p> <div class="markdown-heading" dir="auto"> <h3 class="heading-element" dir="auto" tabindex="-1">🚩 更新</h3> <a id="user-content-triangular_flag_on_post-updates" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#triangular_flag_on_post-updates" aria-label="永久链接：:triangular_flag_on_post: 更新"></a></div> <ul dir="auto"> <li>2025.05.12：🔥🔥 现在支持消费级 GPU（16GB 或 24GB），请参阅<a href="https://github.com/bytedance/DreamO?tab=readme-ov-file#for-consumer-grade-gpus">此处</a>了解说明</li> <li>2025.05.11：🔥🔥我们更新了模型，以缓解过饱和和面部塑形的问题。新版本相比上一版本有了持续的改进。敬请关注！</li> <li>2025.05.08：发布代码和模型</li> <li>2025.04.24：发布DreamO技术报告。</li> </ul> <div class="markdown-heading" dir="auto"> <h2 class="heading-element" dir="auto" tabindex="-1">🔧 依赖项和安装</h2> <a id="user-content-wrench-dependencies-and-installation" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#wrench-dependencies-and-installation" aria-label="永久链接：:wrench: 依赖项和安装"></a></div> <div class="highlight highlight-source-shell notranslate position-relative overflow-auto" dir="auto"> <pre># clone DreamO repo git clone https://github.com/bytedance/DreamO.git cd DreamO # create conda env conda create --name dreamo python=3.10 # activate env conda activate dreamo # install dependent packages pip install -r requirements.txt</pre> <div class="zeroclipboard-container"> </div> </div> <div class="markdown-heading" dir="auto"> <h2 class="heading-element" dir="auto" tabindex="-1">⚡ 快速推理</h2> <a id="user-content-zap-quick-inference" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#zap-quick-inference" aria-label="永久链接：:zap: 快速推理"></a></div> <div class="markdown-heading" dir="auto"> <h3 class="heading-element" dir="auto" tabindex="-1">本地 Gradio 演示</h3> <a id="user-content-local-gradio-demo" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#local-gradio-demo" aria-label="永久链接：本地 Gradio 演示"></a></div> <div class="highlight highlight-source-shell notranslate position-relative overflow-auto" dir="auto"> <pre>python app.py</pre> <div class="zeroclipboard-container"> </div> </div> <p>我们观察到 DreamO 与加速 FLUX LoRA 变体 ( <a href="https://huggingface.co/alimama-creative/FLUX.1-Turbo-Alpha" rel="nofollow">FLUX-turbo</a> ) 之间具有高度兼容性，因此默认启用 Turbo LoRA，将推理步骤减少至 12 步（默认为 25 步以上）。您可以通过 禁用 Turbo <code>--no_turbo</code>，但我们的评估结果好坏参半；因此我们建议保持 Turbo 启用状态。</p> <p>提示：如果观察到肢体扭曲或文本生成不佳，请尝试增加指导比例；如果图像显得过于光泽或过度饱和，请考虑降低指导比例。</p> <div class="markdown-heading" dir="auto"> <p>对于消费级 GPU</p> <a id="user-content-for-consumer-grade-gpus" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#for-consumer-grade-gpus" aria-label="永久链接：适用于消费级 GPU"></a></div> <p>我们添加了对 8 位量化和 CPU 卸载的支持，以便在消费级 GPU 上执行。这需要<code>optimum-quanto</code>库的支持，因此 PyTorch 版本<code>requirements.txt</code>已升级到 2.6.0。如果您使用的是旧版本的 PyTorch，则可能需要重新配置环境。</p> <ul dir="auto"> <li> <p>对于拥有 24GB GPU 的用户，运行<code>python app.py --int8</code>以启用 int8-quantized 模型。</p> </li> <li> <p>对于拥有 16GB GPU 的用户，请运行<code>python app.py --int8 --offload</code>以启用 CPU 卸载和 int8 量化。请注意，CPU 卸载会显著降低推理速度，因此应仅在必要时启用。</p> </li> </ul> <div class="markdown-heading" dir="auto"> <h3 class="heading-element" dir="auto" tabindex="-1">支持的任务</h3> <a id="user-content-supported-tasks" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#supported-tasks" aria-label="永久链接：支持的任务"></a></div> <div class="markdown-heading" dir="auto"> <p>知识产权</p> <a id="user-content-ip" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#ip" aria-label="永久链接：IP"></a></div> <p>此任务与 IP-Adapter 类似，支持多种输入，包括角色、物体和动物。通过利用基于 VAE 的特征编码，DreamO 实现了比以往适配器方法更高的保真度，并在保留角色身份方面具有显著优势。</p> <p><a href="https://private-user-images.githubusercontent.com/11482921/442496266-086ceabd-338b-4fef-ad1f-bab6b30a1160.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NDcxODMwMTQsIm5iZiI6MTc0NzE4MjcxNCwicGF0aCI6Ii8xMTQ4MjkyMS80NDI0OTYyNjYtMDg2Y2VhYmQtMzM4Yi00ZmVmLWFkMWYtYmFiNmIzMGExMTYwLnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNTA1MTQlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjUwNTE0VDAwMzE1NFomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPWY2MWFhODczN2MyZTI3NzA2Zjk2ZmVmYmZjNWRiOWFjYjA3NjVlYjUwZDNmYjVlMzlkZjU4NTE4NTFhMTIwODYmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0In0.QYAzna2IHCyXkpoAlRyCZoojTl6s5miF9rbXajxLIIE" target="_blank" rel="noopener noreferrer"><img style="display: block; margin-left: auto; margin-right: auto;" src="https://private-user-images.githubusercontent.com/11482921/442496266-086ceabd-338b-4fef-ad1f-bab6b30a1160.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.QYAzna2IHCyXkpoAlRyCZoojTl6s5miF9rbXajxLIIE"></a></p> <div class="markdown-heading" dir="auto"> <p>ID</p> <a id="user-content-id" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#id" aria-label="永久链接：ID"></a></div> <p>这里，ID 特指面部身份。与同时考虑人脸和衣着特征的 IP 任务不同，ID 任务仅关注面部特征。该任务与 InstantID 和 PuLID 类似。与之前的方法相比，DreamO 实现了更高的面部保真度，但比 SOTA 方法 PuLID 引入了更多的模型污染。</p> <p><a href="https://private-user-images.githubusercontent.com/11482921/442497094-392dd325-d4f4-4abb-9718-4b16fe7844c6.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.Bf9LLLJjkzVF-iNGorWAwkOCNo1wRapLNeC6x1kG1_4" target="_blank" rel="noopener noreferrer"><img style="display: block; margin-left: auto; margin-right: auto;" src="https://private-user-images.githubusercontent.com/11482921/442497094-392dd325-d4f4-4abb-9718-4b16fe7844c6.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NDcxODMwMTQsIm5iZiI6MTc0NzE4MjcxNCwicGF0aCI6Ii8xMTQ4MjkyMS80NDI0OTcwOTQtMzkyZGQzMjUtZDRmNC00YWJiLTk3MTgtNGIxNmZlNzg0NGM2LnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNTA1MTQlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjUwNTE0VDAwMzE1NFomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPTFkNmM3OGViYzA5NTU1NTlmNzU2M2EwYWQxMmNhOGU3ODc0OWZiMjhkMzYzZmE2MWI0NzI3NDk4MGQ0NzY3YjAmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0In0.Bf9LLLJjkzVF-iNGorWAwkOCNo1wRapLNeC6x1kG1_4"></a></p> <p>提示：如果您发现脸部显得过于有光泽，请尝试降低指导尺度。</p> <div class="markdown-heading" dir="auto"> <p>试穿</p> <a id="user-content-try-on" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#try-on" aria-label="永久链接：试穿"></a></div> <p>此任务支持上衣、下装、眼镜和帽子等输入，并支持多件服装的虚拟试穿。值得注意的是，我们的训练集不包含多件服装或 ID+服装数据，但该模型对这些未见过的组合表现出良好的泛化能力。</p> <p><a href="https://private-user-images.githubusercontent.com/11482921/442498300-fefec673-110a-44f2-83a9-5b779728a734.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.cpTeiHi2QGTk4T7dQ-qTaHoNEsxtwFNZrexo4-noLGE" target="_blank" rel="noopener noreferrer"><img style="display: block; margin-left: auto; margin-right: auto;" src="https://private-user-images.githubusercontent.com/11482921/442498300-fefec673-110a-44f2-83a9-5b779728a734.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.cpTeiHi2QGTk4T7dQ-qTaHoNEsxtwFNZrexo4-noLGE"></a></p> <div class="markdown-heading" dir="auto"> <p>风格</p> <a id="user-content-style" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#style" aria-label="永久链接：风格"></a></div> <p>此任务与 Style-Adapter 和 InstantStyle 类似。请注意，样式一致性目前与其他任务相比不太稳定，并且在当前版本中，样式无法与其他条件组合使用。我们正在努力改进未来版本，敬请期待。</p> <p><a href="https://private-user-images.githubusercontent.com/11482921/442498659-0a31674a-c3c2-451f-91e4-c521659d40f3.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.85KKXNplE4zOXT4KBIj2IKaZ9pqlrxraWAhGvMYpmTo" target="_blank" rel="noopener noreferrer"><img style="display: block; margin-left: auto; margin-right: auto;" src="https://private-user-images.githubusercontent.com/11482921/442498659-0a31674a-c3c2-451f-91e4-c521659d40f3.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.85KKXNplE4zOXT4KBIj2IKaZ9pqlrxraWAhGvMYpmTo"></a></p> <div class="markdown-heading" dir="auto"> <p>多重条件</p> <a id="user-content-multi-condition" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#multi-condition" aria-label="永久链接：多重条件"></a></div> <p>您可以使用多种条件（ID、IP、试穿）来生成更具创意的图像。得益于论文中提出的特征路由约束，DreamO 有效地缓解了多个实体之间的冲突和纠缠。</p> <p><a href="https://private-user-images.githubusercontent.com/11482921/442502721-e43e6ebb-a028-4b29-b76d-3eaa1e69b9c9.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.h6pC7EyXVdKrXyTqoojZRSEKctnIM5o5DhaMFHZcEHM" target="_blank" rel="noopener noreferrer"><img style="display: block; margin-left: auto; margin-right: auto;" src="https://private-user-images.githubusercontent.com/11482921/442502721-e43e6ebb-a028-4b29-b76d-3eaa1e69b9c9.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.h6pC7EyXVdKrXyTqoojZRSEKctnIM5o5DhaMFHZcEHM"></a></p> <div class="markdown-heading" dir="auto"> <h3 class="heading-element" dir="auto" tabindex="-1">在线 HuggingFace 演示</h3> <a id="user-content-online-huggingface-demo" class="anchor" href="https://github.com/bytedance/DreamO?tab=readme-ov-file#online-huggingface-demo" aria-label="永久链接：在线 HuggingFace 演示"></a></div> <p>您可以在<a href="https://huggingface.co/spaces/ByteDance/DreamO" rel="nofollow">HuggingFace</a>上尝试 DreamO 演示。</p>
Search characters for RPers
Manage and collaborate on characters for language models.
