Which one generates a better 3D model from video? Luma AI or 3DPresso?

Olli Huttunen
22 Jul 202316:41

TLDRIn this video, Oligatron compares Luma AI and 3DPresso, two AI-assisted applications for creating 3D models from video. Using a 360 camera, he tests each service on different objects, highlighting the advantages and limitations of each. Luma AI excels in capturing environments and larger subjects, offering real Nerf models suitable for Unreal Engine, while 3DPresso produces more accurate and usable polygon-based mesh models for 3D printing and game assets. The video concludes that both applications have their unique purposes and potential in the evolving field of AI-assisted 3D modeling.

Takeaways

  • 😀 Luma AI and 3DPresso are two AI-assisted applications that generate 3D models from video images.
  • 🔍 Luma AI is developed by Luma Labs, a Silicon Valley startup, while 3DPresso is a Korean alternative.
  • 🌟 Both applications utilize neural Radiance fields (NeRF) technology for 3D modeling.
  • 🎥 The video transcript demonstrates the process of creating 3D models using a 360 camera for scanning different objects.
  • 📸 The 360 camera is particularly useful for scanning larger objects and provides a 180-degree field of view for the front lens.
  • 🖥️ Post-scanning, videos are edited using Insta360 Studio to create a cropped 16x9 format video, which is then fed into the AI programs.
  • ⏱️ Luma AI processes 3D models faster than 3DPresso, but both handle video within a reasonable time frame.
  • 💰 3DPresso uses a credit-based billing structure, while Luma AI remains free to use with some limitations on rendering.
  • 🗿 In the comparison, 3DPresso produced more accurate and complete models for certain objects, especially with better texture quality.
  • 🏺 However, Luma AI excelled in capturing the environment and creating more visually appealing Nerf animations.
  • 🚚 For larger objects like a trailer, Luma AI provided a usable model, whereas 3DPresso failed to produce a viable model.
  • 🔧 Both applications are in early development stages and may not always generate perfect models, indicating the need for further technological advancements.

Q & A

  • What are the two AI-assisted applications mentioned in the script that can create 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 both utilize for creating 3D models from video?

    -Both Luma AI and 3DPresso utilize the neural Radiance field, also known as NeRF technology, for creating 3D models from video.

  • What type of camera does the script recommend for 3D scanning and why?

    -The script recommends using a 360-degree camera for 3D scanning because it can capture a larger area at once, making it suitable for subjects of your own size or larger.

  • How does the script suggest recording video for 3D scanning with a 360 camera?

    -The script suggests recording video in three rounds at different heights, ensuring to capture all angles of the object, and using horizon lock to keep the camera level during rotation.

  • What is the recommended process for editing 360 videos for 3D scanning?

    -The recommended process involves uploading the video to a computer, cropping it to a 16x9 format, adding keyframes to keep the object centered, and exporting it as an mp4 file with a high bitrate for better quality.

  • How does Luma AI differ from 3DPresso in terms of processing time and cost?

    -Luma AI is faster in processing 3D models and is currently free to use, while 3DPresso has a credit-based billing structure with new users receiving free beans to use the service.

  • What feature does Luma AI provide that showcases the 3D model's environment?

    -Luma AI provides a pre-calculated NeRF rendered animation that automatically rotates the camera around the object, showcasing the environment and the model.

  • What is the difference between the Nerf model and the regular polygon model produced by Luma AI?

    -The Nerf model is a pre-calculated rendering that looks different from the regular polygon model when viewed in 3D mode. The polygon model tends to have lumpier surfaces and more holes.

  • How does 3DPresso compare to Luma AI in terms of model accuracy and texture quality?

    -3DPresso produces more accurate models with better texture quality than Luma AI, especially when examined in a web browser or exported in glb format for use in 3D programs like Blender.

  • What are the limitations of Luma AI when it comes to rendering camera animations?

    -Luma AI limits the number of renders to 30 times per month for camera animations, although the service remains free.

  • What conclusion does the script draw about the applications' capabilities for different types of objects?

    -The script concludes that 3DPresso is more suitable for creating accurate and complete models for use in games or 3D printing, while Luma AI is better for capturing environments and larger entities, especially when used with programs like Unreal Engine.

Outlines

00:00

🚀 Introduction to AI-assisted 3D Modeling from Video

The script introduces the concept of AI-assisted 3D modeling from video images, mentioning two key players in the field: Luma AI by Luma Labs and a newer Korean alternative, 3D Presser. Both utilize the trending neural Radiance field (NeRF) technology. The narrator, Oliutron, outlines a plan to compare these services using video files of three different objects captured with a 360-degree camera. The video explains the process of using the camera for 3D scanning, emphasizing the importance of recording from various angles to capture all parts of the object. The script also touches on the challenges of scanning objects with transparent surfaces and the suitability of the 360-degree camera for larger subjects.

05:01

🎥 Post-production and AI Model Processing

This paragraph details the post-production process of the 360-degree videos, including uploading and editing in Insta360 Studio. The video is cropped to a 16x9 format, keyframes are added for object centring, and the videos are exported in mp4 format at 1920x1080 resolution with a high bitrate for quality. The narrator then describes the process of uploading these videos to Luma AI and 3D Presser to see which service can produce the most useful 3D models. It mentions Luma AI's speed advantage and 3D Presser's new credit-based billing system. The paragraph concludes with a comparison of the initial results from both services for a statue model, noting Luma AI's automatic NeRF rendering and 3D Presser's cleaner and more accurate model and textures.

10:02

📐 Detailed Analysis of 3D Models in Blender

The narrator proceeds to examine the statue models within Blender for a more in-depth analysis. Luma AI's model is found to be fragmented with holes, while 3D Presser's model is more compact and less fragmented. The topology of 3D Presser's model follows a grid-like structure, suggesting a better understanding of the object's shape. Despite Luma AI's model having better initial visuals, 3D Presser's model proves to be more accurate and complete, suitable for 3D printing or game assets. The narrator also compares the models of a lamppost and a trailer, noting that Luma AI captures the environment well but struggles with narrow forms, while 3D Presser fails to produce a usable model of the trailer but performs well with the statue and lamppost.

15:04

🔮 Conclusion and Future Prospects of AI 3D Modeling

In the final paragraph, the narrator summarizes the comparison, stating that 3D Presser is more suitable for creating accurate and usable 3D models from video, especially for gaming assets or 3D printing, while Luma AI is better for capturing environments and larger entities. The narrator also acknowledges the early stages of development for these applications and the potential for future advancements. The script concludes by encouraging viewers to test these services themselves and to explore related content on the channel, ending with a call to like, subscribe, and look forward to future videos.

Mindmap

Keywords

💡AI-assisted application

AI-assisted applications are software programs that utilize artificial intelligence to assist users in performing tasks more efficiently. In the context of the video, these applications are used to generate 3D models from video images, showcasing the integration of AI in creative and technical processes.

💡Luma AI

Luma AI is a specific AI-assisted application developed by Luma Labs, a Silicon Valley startup. It is highlighted in the video for its ability to create 3D models from video using neural radiance fields technology, indicating its role in advancing the field of 3D modeling from video content.

💡3DPresso

3DPresso is a Korean alternative to Luma AI, also capable of generating 3D models from video. It is mentioned as a comparative tool to Luma AI, emphasizing the ongoing development and competition in the AI-assisted 3D modeling space.

💡Neural Radiance Fields (Nerf)

Neural Radiance Fields, or Nerf, is a trendy technology used by both Luma AI and 3DPresso to create 3D models. It is a method of representing a 3D scene and is central to the video's theme of comparing different software's capabilities in 3D modeling from video.

💡360 camera

A 360 camera is a device capable of capturing panoramic images or videos. In the video, it is used for 3D scanning by recording videos of objects from all angles, which is essential for creating comprehensive 3D models.

💡Photogrammetry

Photogrammetry is the technique of making measurements from photographs, especially for recovering the exact positions of surface points. In the context of the video, it refers to the challenges of creating 3D models from video, particularly when dealing with transparent or reflective surfaces.

💡Insta360 Studio

Insta360 Studio is an application used for editing 360-degree videos. In the script, it is mentioned as the tool used to process and crop the videos captured by the 360 camera into a format suitable for 3D modeling by AI applications.

💡Polygon-based mesh models

Polygon-based mesh models are 3D models constructed from polygons, the basic building blocks of 3D geometry. The video discusses the quality of these models produced by 3DPresso, noting their accuracy and completeness compared to other methods.

💡Nerf rendering

Nerf rendering is a technique used to create realistic 3D visuals from data. In the video, Luma AI's Nerf rendering is described as a pre-calculated animation that provides an impressive visual representation of the 3D model before further refinement.

💡3D scanning

3D scanning is the process of analyzing a real-world object to collect data on its shape and possibly its appearance. The video demonstrates the use of a 360 camera for 3D scanning, which is a key step before using AI applications to generate 3D models.

💡GLB format

GLB is a file format for 3D models that is optimized for the web and supports embedded textures. The video mentions exporting models in GLB format for further examination in 3D modeling software like Blender, highlighting the importance of this format in the workflow of 3D modeling from video.

Highlights

Luma AI and 3DPresso are two AI-assisted applications that generate 3D models from video images.

Luma AI is developed by Silicon Valley startup Luma Labs, while 3DPresso is a Korean alternative.

Both applications use neural Radiance fields, also known as NeRF technology, for 3D modeling.

Olioutron compares Luma AI and 3DPresso by testing them with 360 camera video files of different objects.

The 360 camera is utilized to capture 180-degree views suitable for 3D scanning.

A selfie stick is recommended for recording 360 videos to avoid appearing in the final scan.

The camera's Horizon lock feature ensures level recording even when the selfie stick is turned.

Three rounds of recording at different heights are recommended for capturing the object fully.

Luma AI processes 3D models faster than 3DPresso but both handle one video within 30 minutes.

3DPresso introduces a credit-based billing structure with free beans for new users.

Luma AI remains free to use with some restrictions on rendering camera animations.

Luma AI's pre-calculated NeRF rendering creates impressive animations but may differ in 3D mode.

3DPresso produces more accurate models and better textures compared to Luma AI in web browser views.

Both Luma AI and 3DPresso have limitations, with Luma's model having fragmented textures and 3DPresso struggling with larger objects.

3DPresso's models are more accurate and complete, making them better for 3D printing and game assets.

Luma AI excels in capturing environments and larger entities, suitable for use with programs like Unreal Engine.

The technology behind both applications is rapidly developing, with potential yet to be fully realized.

Olioutron encourages viewers to test these services themselves, noting the ease of video production with smartphones.