Which one generates a better 3D model from video? Luma AI or 3DPresso?
TLDRIn this video, the creator compares two AI-assisted applications, Luma AI and 3DPresso, for generating 3D models from video images. Both utilize neural radiance fields (NeRF) technology, but the video explores which one produces better results. The creator uses a 360 camera to capture videos of three different objects and then processes them through both services. Luma AI is faster but has limitations on rendering, while 3DPresso offers a credit-based billing system. The video showcases the strengths and weaknesses of each application, noting that 3DPresso produces more accurate and complete models, making it better for 3D printing or game assets. However, Luma AI excels at capturing environments and larger entities, making it ideal for use with engines like Unreal Engine. The video concludes by acknowledging the rapid development in this technology and encourages viewers to experiment with these services themselves.
Takeaways
- 🚀 Luma AI and 3DPresso are two AI-assisted applications that can generate 3D models from video images.
- 🌟 Both applications use neural Radiance fields (NeRF) technology to create 3D models.
- 📸 The video demonstrates the use of a 360 camera for capturing video of objects for 3D scanning.
- 🔍 Luma AI processes 3D models faster than 3DPresso, but both handle one video within approximately 30 minutes.
- 💰 3DPresso has a credit-based billing structure, while Luma AI remains free to use with some rendering limitations.
- 🎉 Luma AI automatically creates a NeRF-rendered animation, offering a 'wow' factor for viewers.
- 📐 3DPresso produces more accurate and complete polygon-based mesh models than Luma AI.
- 🧩 Luma AI's models have a fragmented structure with many holes, requiring significant editing for watertight models.
- 🏺 3DPresso handles transparent glass surfaces better than Luma AI, but both struggle with very narrow parts.
- 🚫 3DPresso failed to produce a usable model for a larger object, whereas Luma AI succeeded.
- 🔮 NeRF models from Luma AI are best utilized with programs like Unreal Engine that understand volume models.
- ⏱ As the technology develops, the potential use cases for NeRF models are expected to expand.
Q & A
What are the two AI-assisted applications mentioned for generating 3D models from video images?
-The two AI-assisted applications mentioned are Luma AI, developed by Luma Labs, and 3DPresso, a Korean alternative.
What technology do Luma AI and 3DPresso use to create 3D models from video?
-Both Luma AI and 3DPresso use a neural Radiance field, also known as NeRF technology, to create 3D models from video.
What is the advantage of using a 360 camera for 3D scanning?
-A 360 camera is advantageous for 3D scanning because it can capture larger areas at once, making it suitable for scanning objects of human size or larger.
How does the Horizon lock feature on the 360 camera help during the scanning process?
-The Horizon lock feature keeps the camera leveled even when the selfie stick is turned, ensuring that the person operating the camera does not appear in the final image.
What are the keyframe settings used for processing the 360 videos in the Insta360 Studio application?
-The keyframe settings used are a value of 50 for the focal length and 15 for the lens Distortion setting.
How does Luma AI differ from 3DPresso in terms of processing time?
-Luma AI is currently faster at processing 3D models compared to 3DPresso.
What is the new billing structure introduced by 3DPresso?
-3DPresso has introduced a credit-based billing structure called the 'beam system' which controls and charts their services on the website.
What are the limitations of Luma AI when it comes to rendering camera animations?
-Luma AI limits the renders to 30 times per month for camera animations.
How does Luma AI handle the background in the generated 3D models?
-Luma AI manages to capture the background around the object, although the model may not look as accurate when viewed in 3D mode.
What is the main difference between the 3D mesh quality of models produced by Luma AI and 3DPresso?
-3DPresso produces more accurate and complete models with better single image textures, while Luma AI's models are less accurate with lumpy surfaces and holes.
Which application is more suitable for creating accurate 3D assets for games or 3D printing?
-3DPresso is more suitable for creating accurate 3D assets for games or 3D printing, while Luma AI is better for saving environments and larger entities.
What is the current limitation of these AI applications in terms of model creation?
-These applications are in their early stages of development and do not always manage to form a working model every time, especially for complex or challenging objects.
Outlines
🚀 Introduction to AI-assisted 3D Modeling from Video
The video begins by discussing the scarcity of AI applications capable of creating 3D models from video images, mentioning two such services: Luma AI by Luma Labs and a newer Korean alternative called 3D Presser. Both utilize neural radiance fields, also known as NeRF technology. The host, Oliutron, introduces their 3D scanning journey with these AI-listed softwares, comparing their capabilities to produce useful 3D models from video files. The process involves using a 360-degree camera to capture three different objects, with a focus on the technique's principles and the advantages of using a 360-degree camera for scanning larger subjects. The video also covers the recording process, including the use of a selfie stick and the importance of capturing all angles of the object for a complete 3D model.
🎥 Post-Processing and AI Model Generation
The second paragraph describes the post-processing of the 360-degree video footage. The host transfers the material to a computer and uses the Insta360 Studio application to crop the video into a 16x9 format, adding keyframes to keep the object centered. The video is then exported as an mp4 file with a high bitrate for quality. The final phase involves uploading the video files to Luma AI and 3D Presser to generate 3D models. The host compares the processing speed and cost of both services, noting that Luma AI is currently faster and free, while 3D Presser has a credit-based billing system. The results from both services are then examined, with a focus on the statue model. Luma AI's model is less accurate with lumpier surfaces and holes, whereas 3D Presser's model is more accurate and cleaner, with better textures.
🏆 Evaluating Model Quality and Usability
The third paragraph delves into the evaluation of the 3D models within Blender, a 3D modeling software. The host highlights that 3D Presser's model is more intact and less fragmented, making it more suitable for uses such as 3D printing. However, Luma AI's model, despite its inaccuracies, captures the surroundings well. When examining the lamppost model, both services struggle with the transparent parts and narrow details, but 3D Presser performs slightly better. In the case of the trailer, 3D Presser fails to produce a usable model, while Luma AI creates a good model that captures reflective properties effectively. The host concludes that each application has its purpose; 3D Presser is better for accurate models, and Luma AI is ideal for capturing environments and larger entities.
🔮 Future Prospects and Conclusion
The final paragraph discusses the future potential of NeRF technology and the development path of Luma AI and similar applications. The host expresses excitement about the direction these technologies are taking and encourages viewers to experiment with the services using their smartphones or 360 cameras. The video concludes with a reminder to like and subscribe for more content on 3D scanning and related topics.
Mindmap
Keywords
💡AI-assisted application
💡Luma AI
💡3DPresso
💡Neural Radiance Fields (NeRF)
💡360 Camera
💡Photogrammetry
💡Insta360 Studio
💡Polygon Model
💡GLB Format
💡Blender
💡Nerf Rendered Animation
Highlights
Luma AI and 3DPresso are AI-assisted applications that can create 3D models from video images.
Luma AI is developed by Luma Labs, a Silicon Valley startup, while 3DPresso is a Korean alternative.
Both applications use neural Radiance fields (NeRF) technology for 3D model generation.
Olioutron, the presenter, uses a 360 camera to capture videos for 3D scanning.
The 360 camera is particularly useful for scanning larger objects due to its wide field of view.
Luma AI processes 3D models faster than 3DPresso.
3DPresso has introduced a credit-based billing structure called 'beam system'.
Luma AI is currently free to use, with some rendering limitations.
3DPresso provides new users with 100 free beans to start using the service.
Luma AI automatically creates a NeRF-rendered animation for each model.
3DPresso's models have better accuracy and texture quality compared to Luma AI.
3DPresso's models are more suitable for 3D printing and game development purposes.
Luma AI is better for capturing environments and larger entities due to its NeRF capabilities.
Luma AI's models are more fragmented and require more work to be watertight for 3D printing.
3DPresso's polygonal structure is more tense and compact, making it more usable.
Both applications struggle with very narrow or transparent parts in objects.
3DPresso failed to produce a usable model for a larger object, while Luma AI succeeded.
These applications are in early development stages and do not always form a working model.
3DPresso is currently better for producing usable 3D models from videos, but the technology is rapidly evolving.
The presenter encourages viewers to test these services themselves, noting that any camera capable of recording video can be used for 3D scanning.