-
Notifications
You must be signed in to change notification settings - Fork 6
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
37f5a93
commit b719da7
Showing
2 changed files
with
102 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,101 @@ | ||
--- | ||
title: 7 月 12 日 | ||
date: 2024-07-12 | ||
cover: https://oss.justin3go.com/blogs/fav0-001.jpg | ||
|
||
--- | ||
|
||
每天仅需 1 分钟,全面获取 AI 技术发展、行业动态和市场趋势。 | ||
|
||
内容涵盖但不限于**前沿 AI 资讯**、**AI 工具**、**AI 绘画**、**开源项目**和**学习教程**等等。 | ||
|
||
关注 AI 日报,紧跟 AI 潮流,希望对你有所帮助。对于重要信息,会独立发帖进行详细介绍。 | ||
|
||
以下是 7 月 12 日的最新 AI 信息。 | ||
|
||
### 前沿资讯 | ||
|
||
**1、音频生成视频项目实在太卷了,HeyGen 也加入战场!** | ||
|
||
推出 Expressive Photos 功能,类似于阿里的 Emo,输入角色图片和音频,为角色生成逼真且嘴型匹配的视频。 | ||
|
||
而且,带有面部表情和头部动作。目前,可以在 HeyGen 上免费体验。 | ||
|
||
体验地址:https://labs.heygen.com/expressive-photo-avatar | ||
|
||
![image-20240712155250929](https://cdn.jsdelivr.net/gh/freelander/oss@master/ai-daily/2024-07-12/image-20240712155250929.png) | ||
|
||
|
||
|
||
### 前沿技术 | ||
|
||
**1、一个开源的原生图文多模态模型 SOLO。** | ||
|
||
与目前 GPT-4o 类似,使用的是统一的 Transformer 架构,能接受图片和文字输入,输出文字。 | ||
|
||
从给出的数据来看,SOLO-7B 与 LLaVA-v1.5-7B 能力接近,LLaVA 是 LLM+ 视觉编码器组合而成的多模态模型。 | ||
|
||
GitHub:https://github.com/Yangyi-Chen/SOLO | ||
|
||
模型下载:https://huggingface.co/YangyiYY/SOLO-7B | ||
|
||
![fig1.png](https://cdn.jsdelivr.net/gh/freelander/oss@master/ai-daily/2024-07-12/fig1.png) | ||
|
||
**2、利用多模态语言模型构建生动的图文故事 SEED-Story。** | ||
|
||
一种创新的方法,利用多模态大语言模型生成丰富文本和上下文相关图像的长篇图文故事。 | ||
|
||
而且可以保持整体故事下来的图像角色一致性,同时能够有效生成长篇高质量的故事。 | ||
|
||
GitHub:https://github.com/TencentARC/SEED-Story | ||
|
||
论文地址:https://arxiv.org/abs/2407.08683 | ||
|
||
![Teaser image](https://cdn.jsdelivr.net/gh/freelander/oss@master/ai-daily/2024-07-12/teaser.jpg) | ||
|
||
|
||
|
||
### AI 绘画 | ||
|
||
**1、ComfyUI 中的 3D 插件 ComfyUI-3D-Pack 迎来重磅更新!** | ||
|
||
ComfyUI-3D-Pack 插件,旨在让你在 ComfyUI 中生成 3D 资源就像生成图像/视频一样简单便捷。 | ||
|
||
- 现在支持从 ComfyUI-Manager 一键安装。 | ||
|
||
- 集成了 CharacterGen 管道,并对 Unique3D 管道进行改进。 | ||
|
||
GitHub:https://github.com/MrForExample/ComfyUI-3D-Pack | ||
|
||
CharacterGen 是一个高效的 3D 角色生成框架,能够从单张输入图像生成具有统一姿势的 3D 角色网格。 | ||
|
||
![Comfy3D Update](https://cdn.jsdelivr.net/gh/freelander/oss@master/ai-daily/2024-07-12/Comfy3D%20Update.gif) | ||
|
||
|
||
|
||
### 开源项目 | ||
|
||
**1、一款开源的本地大模型 Mac 客户端 Enchanted。** | ||
|
||
类似于 ChatGPT 客户端,需搭配 Ollama 使用,轻松连接本地部署的私有模型,如 Llama2、Mistral 以及 Vicuna 等。 | ||
|
||
并且,提供了在 iOS 生态系统(macOS、iOS、Watch、Vision Pro)中无过滤、安全、私密和多模态的使用体验。 | ||
|
||
GitHub:https://github.com/AugustDev/enchanted | ||
|
||
![img](https://cdn.jsdelivr.net/gh/freelander/oss@master/ai-daily/2024-07-12/1&e=1725119999&s=mtvyvvtvyyyj&token=kIxbL07-8jAj8w1n4s9zv64FuZZNEATmlU_Vm6zD:6YSEy-HluS2NcnEIp5mJQa6iOU0=.png) | ||
|
||
|
||
|
||
**2、一个可在浏览器内使用的语音识别 Whisper Timestamped,带有单词级时间戳。** | ||
|
||
基于 Transformers.js,可在浏览器中本地运行使用,自动下载模型 whisper-base(带时间戳),能够识别 100 种不同语言并生成单词级时间戳。 | ||
|
||
模型大小仅 196MB,首次加载完成后,即使后面在断网情况下,依然能重复使用。 | ||
|
||
GitHub:https://github.com/xenova/transformers.js/tree/v3/examples/whisper-word-timestamps | ||
|
||
在线体验:https://huggingface.co/spaces/Xenova/whisper-word-level-timestamps | ||
|
||
![img](https://cdn.jsdelivr.net/gh/freelander/oss@master/ai-daily/2024-07-12/1&e=1725119999&s=mtvyvvtvyyyj&token=kIxbL07-8jAj8w1n4s9zv64FuZZNEATmlU_Vm6zD:omFKQKHjOsCi6DqphIa_86Ke_vw=.png) | ||
|