Xunguang

• 1. overview (675 chars)

  Xunguang is a one-stop AI video creation platform developed by Alibaba Damo Academy and was first released at the 2024 Shanghai World Artificial Intelligence Conference (WAIC). The platform provides full-process services from script creation to storyboard design, character customization, scene generation and video editing, and is positioned as a PUGC (professional user-generated content) tool. Different from traditional Vincent video tools, Xunguang focuses on "precisely controlled editing" of video content and supports core functions such as AI lip synchronization, character posture control, intelligent target elimination and artistic style transfer. The product is currently in the free internal testing stage, and you will receive 100 light points every day when you log in.

• 2. Product background (894 chars)

  Alibaba Damo Academy Vision Technology Laboratory is the R&D team behind Xunguang AI. Founded in 2017, DAMO Academy is an institution under the Alibaba Group that focuses on basic research and disruptive technological innovation. It has profound technological accumulation in the fields of artificial intelligence, machine learning, and computer vision. Xunguang AI will be officially released at the WAIC 2024 conference in July 2024. It is an important layout of DAMO Academy in the field of AI video creation. Relying on the multi-modal large model and 3D face deformation technology independently developed by DAMO Academy, Xunguang has unique technical advantages in precise video control.   From the perspective of product positioning, Xunguang is not a simple Vincent video tool, but a "precisely controlled video editing platform." This positioning enables it to form differentiated competition with Wensheng video tools such as Runway and Pika, and focuses more on the post-production needs of professional video creators and corporate teams.

• 3. Product functions and usage experience (1954 chars)

  The core functions of Xunguang AI can be divided into four major modules, each of which reflects DAMO Academy’s technological accumulation in the field of computer vision.   The first is the intelligent AI mouth shape driving function. This function is based on the multi-modal model self-developed by DAMO Academy and supports users to upload text or local dubbing. The system generates natural and accurate lip-sync animations through millisecond-level facial key point matching. This function is particularly useful for short video creators to produce narration videos, which can significantly reduce the time cost of dubbing and lip matching.   Secondly, there are high-precision character and posture control functions. Through the 3D face deformation model and bone binding technology, users can finely control the expressions and postures of the characters in the video. The platform presets a variety of expressions and action combinations, and creators can flexibly choose and apply them to videos. This function has high practical value in the fields of film and television pre-research and game animation design.   The third is the intelligent target elimination and repair function. Using advanced image repair algorithms, users can remove unnecessary debris or pedestrians from the video with one click, and the system will automatically complete the background to achieve traceless screen cleaning. This function is similar to Inpaint technology in the picture field, but is specially optimized for video scenes.   The fourth is the multi-dimensional art style transfer function. The platform has built-in professional-level visual styles such as ink and wash, Xin Haicheng, etc., and combined with a lightweight rendering engine, users can quickly complete the artistic reshaping of videos. This feature provides video creators with rich means of creative expression.   In addition to the above four core functions, Xunguang also provides auxiliary functions such as storyboard generation, character library management, visual material creation, video content editing, camera movement control and motion control, foreground generation and layer editing, forming a complete video creation workflow. Users can complete all creative aspects from script to finished film on the platform.

• 4. Pricing and business model (900 chars)

  Xunguang AI is currently in the free internal testing stage, and no specific commercial pricing plan has yet been announced. According to the information on the product page, users who log in every day can get 100 light points as a free quota, and generating an AI video consumes about 10 light points. This model is similar to the membership subscription system of ByteDance's Clip Movies, which attracts user experience by providing free credits, and then realizes commercial monetization through value-added services.   Judging from industry practice, the charging models of AI video creation platforms usually include charging based on the duration of production, charging based on membership subscription, and charging based on light points/points. The light spot system selected by light search is a relatively flexible billing method that users can use flexibly according to actual needs. For enterprise users, the platform may launch enterprise-level subscription plans in the future, providing higher generation quotas and priority processing queues.

• 5. User reviews (732 chars)

  Since the product is still in the internal testing stage, user reviews from public channels are relatively limited. Judging from reports in the product community and technology media, Xunguang AI's lip synchronization technology and character posture control functions have received high attention, and the industry remains hopeful about its technical strength.   From the perspective of product positioning, Xunguang’s target user groups include short video creators, self-media practitioners, corporate marketing departments, educational institutions, film and television game production teams, etc. The platform's full-process creation capabilities and precise control features make it more suitable for user groups with professional video production needs, rather than Wensheng video tools that are completely oriented to ordinary consumers.

• 6. Industry evaluation (685 chars)

  In the field of AI video creation, the release of Xunguang AI is regarded as Alibaba’s important layout in the field of generative AI. Compared with domestic competitors such as ByteDance’s Cutout AI and Baidu’s Dujia creation tool, Xunguang’s differentiated advantage lies in its product concept of “precise controlled editing” and the technical endorsement of DAMO Academy.   From the perspective of the industry landscape, competition in the AI ​​video creation track is becoming increasingly fierce. In the international market, products such as Runway, Pika, and OpenAI's Sora remain in the lead, while many companies are present in the domestic market. If Xunguang wants to stand out from the competition, it needs to continue to optimize product experience, production quality, and commercialization.

• 7. Controversies and Risks (602 chars)

  At present, the public controversy about Xunguang AI is relatively limited. Potential risks that require attention include:   One is copyright and legal risks. AI video generation involves the use of materials and the copyright ownership of generated content. The platform needs to clarify the rights boundaries of user-generated content.   The second is the risk of technology abuse. Precise video editing tools may be used for improper purposes such as deepfakes, and platforms need to establish corresponding review and protection mechanisms.   The third is the challenge of commercialization. After the free internal beta phase, how to balance user experience and commercial benefits is one of the challenges.

• 8. Applicable groups and usage suggestions (654 chars)

  Xunguang AI is suitable for the following user groups: short video creators and self-media practitioners can use the lip synchronization function to quickly produce explanation videos; corporate marketing departments can be used for product demonstrations and advertising production; educational institutions can be used for teaching animation and AI digital human explanations; film and television game teams can be used for pre-research and storyboard design.   For ordinary consumers, if it is mainly used for simple video needs, they can consider international competing products such as Runway and Pika or domestic tools such as AI. If you need precise video control and professional-level post-production capabilities, light hunting is an option worth trying.

• 9. Summarize (423 chars)

  Xunguang AI is an important product of Alibaba DAMO Academy in the field of AI video creation. With its differentiated positioning of "precise controlled editing" and DAMO Academy's technology accumulation, it has demonstrated unique competitiveness in the fiercely competitive AI video track. The product is currently in the free internal testing stage. It is recommended that interested users experience it in advance and pay attention to the official commercialization process of the product.

No in-depth report yet

• 

  CHgra

  —

  刚体验了寻光AI的口型同步功能，确实挺香的！比我之前用的那些工具精准太多了，而且免费额度也够用。

---

• 

  fxym1auo1

  —

  阿里达摩院出品必属精品！视频消除功能太实用了，一键去掉路人和水印，效果还很自然，栓Q了！

---

• 

  DrNinelPantelyuk

  —

  个人感觉风格迁移是亮点，水墨和新海诚风格效果不错，适合做国风内容。

---

• 

  سپهرکوتی

  —

  正在内测中，整体体验还行，就是希望能开放更多角色模板。

---

• 

  6w43g_n

  —

  用了两周，感觉比Runway更适合做精准控制的视频，国产之光！

---

• 

  Hannah_Turner_20238

  —

  说实话免费额度有点少，每天100光点很快就用完了，期待正式版的价格出来。

---

• 

  AltSeasonJimenez

  —

  3D姿态控制 yyds！终于可以自己定义角色动作了，不用受限于预设动画。

---

• 

  ALcox_fi

  —

  达摩院的技术确实强，生成的视频质量很高，期待更多功能上线！

---

• 

  SamuelPatelSr04

  —

  作为一个短视频博主，寻光真的帮我省了很多后期时间，AI消除路人 yyds！

---

• 

  EReyes_202004

  —

  和剪映对比了一下，寻光在专业视频编辑方面更强，适合有一定基础的用户。

---

• 

  George_TurnerIII

  —

  期待正式版！目前内测阶段已经这么好用了，以后正式上线还得了。

---

• 

  JasonWilliams_2021

  —

  运镜控制功能很强大，可以精细调整镜头运动轨迹，这个功能我爱了！

---

• 

  k68t3v4q5k

  —

  故事板生成功能救了我的命！以前要花几个小时想分镜，现在几分钟就搞定了。

---

• 

  Sara657

  —

  希望官方能出移动端 App，这样出门也能编辑视频就更方便了。

---

• 

  Barbara_HowardK

  —

  阿里达摩院牛批！多模态大模型技术确实领先，生成的视频质感很棒。

---

• 

  程瑶浩

  —

  目前内测阶段功能已经这么完善了，正式版值得期待！

---

• 

  NGarcia369

  —

  用了几天感觉不错，特别是口型同步，几乎没有延迟，效果很自然。

---

• 

  Alan.Price

  —

  作为影视从业者，个人很看好寻光在预研阶段的潜力，能快速出分镜和概念视频。

---

• 

  KarolineSandvold

  —

  希望能开放API接口，这样就可以集成到自己的工作流里了。

No reviews yet