腾讯混元视频:ComfyUI官方原生支持!GGUF量化低显存福音!开源生态高速发展

科技   2024-12-20 06:54   浙江  

腾讯混元视频:ComfyUI官方原生支持!GGUF量化更低显存!

🌹大家好!欢迎来到破狼公众号。感谢大家的支持与鼓励。在AIGC探索道路上,我将与你一路同行。喜欢就星标关注破狼公众号或文末扫码加入交流群 !

HunyuanVideo GGUF简介

在之前文章已经介绍过腾讯最新开源的130亿参数的文生视频大模型Hunyuan Video(混元视频)。是当前AI视频开源社区最具代表性先进视频模型之一,在很多方面能够足够媲美OpenAI最新开放的Sora视频模型。目前社区大佬也在发力改善腾讯混元视频生态,已有LORA微调生态支持、图片风格参考IP2VFP8和GGUF量化模型支持等。特别在近日,混元视频也迎来了ComfyUI软件官方的原生支持,现在能够在ComfyUI中轻松实现混元视频生成的功能。

HunyuanVideo GGUF ComfyUI体验

当前ComfyUI官方已原生支持HunyuanVideo的体验了,仅需更新ComfyUI软件本体到最新版本即可。下面所需模型均可文末网盘下载

  • • hunyuan_video_t2v_720p_bf16.safetensors :下载hunyuan_video_t2v_720p_bf16.safetensors文件,并将其放置在ComfyUI的models/unet文件夹中。下载地址:https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/tree/main/split_files/diffusion_models
  • • clip_l和llava_llama3_fp8_scaled:下载clip_l.safetensorsllava_llama3_fp8_scaled.safetensors文件,并将其放置在ComfyUI的models/clip目录中。下载地址:https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/tree/main/split_files/text_encoders
  • • hunyuan_video_vae_bf16.safetensors:还需下载hunyuan_video_vae_bf16.safetensors文件,并将其放置在ComfyUI的models/vae文件夹中。下载地址:https://huggingface.co/Comfy-Org/HunyuanVideo_repackaged/tree/main/split_files/vae
  • • 如果使用GGUF量化版本可以配合ComfyUI-GGUF插件与ComfyUI原生工作流一起使用,能够显著降低显存要求。GGUF模型下载:https://huggingface.co/city96/HunyuanVideo-gguf/tree/main
  • • 和之前ComfyUI-HunyuanVideoWrapper插件本次新增的模型为llava_llama3_fp8_scaled模型。

Flux文生图工作流

Flux文生图感兴趣的同学可参考LIBLIB在线运行工作流:FLUX[续篇]:12B参数23G最大开源文生图模型,Dev版直出惊艳美图欣赏
本文涉及ComfyUI工作流和模型均可在LIBLIBAI上下载或在线运行体验:

• F.1-绮梦流光-水湄凝香

https://www.liblib.art/modelinfo/134c6dd95aef48e98a22b24e003e026b

• 工作流-Flux文|图生图+LORA+提示反推一键切换工作流

https://www.liblib.art/modelinfo/782aacd70f604da39e83368c696a02a8


HunyuanVideo GGUF视频工作流

HunyuanVideo视频工作流已上传LIBLIB平台

https://www.liblib.art/modelinfo/433fbf0bd2a8484d8e32d9e32258f378?versionUuid=aba77a6852f24c209fc1c5a4bf520505

注意

  • • ComfyUI官方工作流仅需20步生成视频,无需原始的50步。同时GGUF视频质量保持很好。

  • • 可以仅需替换Unet加载为GGUF Unet加载器即可。对于显存使用GGUF量化模型12G显存可运行。24G显存2秒 720*480分辨率视频仅需约2分钟5秒视频大约5分钟左右

  • • 预测12G显存理论可体验混元大模型啦,更低8G显存需要待测试反馈。


01. 演讲

cinematic, anime aesthetic,The camera is fixed in a carefully arranged study room, where kxsr is sitting at his desk facing the audience, wearing traditional Chinese medicine clothes and carrying on his back a bookshelf lined up with ancient books and medicine cabinets. His eyes reveal deep wisdom. Facing the camera, he tells the story of traditional Chinese medicine in a gentle and firm voice. The picture is presented in high quality, with calm colors and soft light and shadow, creating a solemn and warm atmosphere.

02.动漫女孩

anime style anime girl with massive fennec ears and one big fluffy tail, she has blonde hair long hair blue eyes wearing a pink sweater and a long blue skirt walking in a beautiful outdoor scenery with snow mountains in the background

03.冬天

cinematic, anime aesthetic,A young Chinese woman, wearing a white down jacket and a white scarf, looks directly at the audience as she walks down the street in Harbin, watching the snow fall around her

混元视频模型:关注公众号口令【视频模型混元目录获取
更多推荐文章:
• [ComfyUI]MagicQuill:无需PS全功能P图编辑工具!8G可运行,最全避坑指南
• [ComfyUI]腾讯混元视频:运动镜头控制!行走动画LORA,最强开源视频可控性里程碑
• [ComfyUI]腾讯混元视频:开源LORA生态来啦!视频人物一致性&运镜可控性里程碑
• [ComfyUI]OminiControl:Flux全能P图神器!乾坤大挪移&重绘&CN控制,极简全能控制器
• [ComfyUI]Flux:国潮经典!敦煌古韵与现代极简国潮插画,敦煌风十二生肖和邮票集锦
• [ComfyUI]Flux:MJ艺术风!基于黑森林F1的Midjourney风,光效&写实&建模&演示&特写镜头
感兴趣加入[AGI技术交流群]+V

    如果觉得文章不错,就请在看转发三连

破狼
关注AIGC、LLM、绘图作品、软件工程、技术学习。交流+V:shunshizhiwu。
 最新文章