? 请关注我获取最新更新 https://twitter.com/camenduru ? 请加入我们的discord服务器 https://discord.gg/k5BwmmvJJU

Potat 1️⃣

首个开源1024x576文字到视频模型 ?

https://huggingface.co/vdo/potat1-5000/tree/main https://huggingface.co/vdo/potat1-10000/tree/main https://huggingface.co/vdo/potat1-10000-base-text-encoder/tree/main https://huggingface.co/vdo/potat1-15000/tree/main https://huggingface.co/vdo/potat1-20000/tree/main https://huggingface.co/vdo/potat1-25000/tree/main https://huggingface.co/vdo/potat1-30000/tree/main https://huggingface.co/vdo/potat1-35000/tree/main https://huggingface.co/vdo/potat1-40000/tree/main https://huggingface.co/vdo/potat1-45000/tree/main https://huggingface.co/vdo/potat1-50000/tree/main https://huggingface.co/vdo/potat1-50000-base-text-encoder/tree/main = https://huggingface.co/camenduru/potat1 (你在这里)

信息

原型模型 使用1个A100（40GB）进行训练 2197个剪辑，68388个标记帧 ( salesforce/blip2-opt-6.7b-coco ) train_steps: 10000

数据集和配置

https://huggingface.co/camenduru/potat1_dataset/tree/main

微调

https://github.com/Breakthrough/PySceneDetect https://github.com/ExponentialML/Video-BLIP2-Preprocessor https://github.com/ExponentialML/Text-To-Video-Finetuning https://github.com/camenduru/Text-To-Video-Finetuning-colab

基础模型

https://huggingface.co/damo-vilab/modelscope-damo-text-to-video-synthesis https://www.modelscope.cn/models/damo/text-to-video-synthesis

感谢 damo-vilab ❤ ExponentialML ❤ kabachuha ❤ @DiffusersLib ❤ @LambdaAPI ❤ @cerspense ❤ @CiaraRowles1 ❤ @p1atdev_art ❤

感谢Orellius ❤（重要错误报告）

请尝试一下 ? https://github.com/camenduru/text-to-video-synthesis-colab

Potat 2️⃣ 在烤箱中 ♨