Immersity AI
Leia Inc. 出品的云端 2D 转 3D 工具，用神经深度引擎把平面图片视频加上深度与运动，导出 Apple Vision Pro 等 XR 设备可看的沉浸式内容
In-depth Report
-
Immersity AI (formerly LeiaPix Converter) is a cloud-based "2D to 3D" tool launched by Leia Inc. It uses a self-developed neural depth engine to add depth and motion to flat images and videos, and export them into immersive content that can be viewed on XR devices such as Apple Vision Pro and Meta Quest. Its most tangible differences are two points: first, it directly connects to Vision Pro’s MV-HEVC spatial video format, and second, it can create dynamic album covers for Apple Music. But you have to accept its payment logic - the free file is watermarked and only available in 720p. If you really want to use it, you have to buy credits. The cost for heavy video users will increase significantly.
-
The company is called Leia Inc. and was founded by David Fattal between 2013 and 2014. He was originally a physicist from Stanford and did diffraction light field research at HP Labs. He was inspired by the light scattering he saw in the parking lot during a fire drill. The company's name paid tribute to the holographic image of Leia in "Star Wars". In the early days, Leia made naked-eye 3D light field display hardware, such as making a 3D screen for the RED Hydrogen One mobile phone and launching the Lume Pad immersive tablet. In 2023, it acquired the 3D patent portfolios of Dimenco and Philips. With more than 2,000 patents in hand, it claims to be No. 1 in the immersive display industry. By 2024, the company shifted its focus from hardware to AI, and the product was renamed Immersity AI; at the end of the same year, it integrated the dynamic cover function of Apple Music; on March 26, 2025, Immersity 4.0 was released, with more natural depth rendering, and a free iOS / Android mobile app was launched. The company is headquartered in Menlo Park, USA, with offices in California, the Netherlands, and China, with more than 200 employees, and claims to have more than 5 million users (third-party agencies claim more than 3 million).
-
The usage process is very simple: upload materials in JPG, PNG, MP4, GIF, WEBP, MKV, MOV and other formats, the engine will automatically generate a depth map, split the foreground and background into different layers, and then use inpainting to complete the occluded details. Users can choose from 11 motion styles, including horizontal, vertical, circular, perspective, zoom, dolly, etc. They can also adjust the motion intensity, focal length and camera path, and preview in real time before exporting if they are satisfied. Export formats cover MP4, GIF, PNG, SBS left and right formats (for Meta Quest and older 3D TVs), MV-HEVC (for Vision Pro's spatial video), and Apple Music dynamic covers (3:4, 1:1 with blurred background). The depth map itself can also be exported separately and used for further synthesis in Blender, After Effects, and Photoshop. It supports batch processing of up to 5 images or videos at a time; the free mobile app is watermarked and is for non-commercial use only; the web paid version supports 4K, commercial use, and no watermark; there are also APIs for developers and embedded solutions for OEMs.
-
New users who sign up will receive 100 credits for trial use. The free file is 720p, watermarked, and non-commercial. The official fee is based on a credit point system: the picture package starts at $4.99 per month for 500 credits, and the maximum is $14.99 for 1,700 points; the video package ranges from $29.99 per month (3,300 points) to $99.99 (20,000 points); in addition, one-time credit packages are sold, ranging from $5 to $100, corresponding to 500 to 12,000 points. A single premium export costs about 50 credits, videos are billed by the second, and 4K is quite a bit more expensive than 1080p. There is also a Plus subscription for $10 per month, which includes 1,000 monthly credits, priority rendering, and commercial licensing.
-
The score for tool directory sites is approximately between 4.0 and 4.5, but the sample size is small and the reference value is limited. The real reviews that can be found are generally positive. What everyone likes the most is the "zero threshold" and the depth effect-some users threw in travel sunset photos, and it came out like the opening of a movie blockbuster. The negative voices are almost all concentrated on the paid version: the free version's watermark plus 720p "cannot be distributed at all", the credit points of long videos are burned quickly, and complex backgrounds occasionally appear edge distortion (warping). The editor of an evaluation media gave it a score of 3.5, concluding that "it does one thing right, and only one thing." Although the words are harsh, they point out its positioning.
-
On the narrow track of "2D to 3D depth animation", Immersity AI is generally regarded as the leader, especially since it has become a rare civilian pipeline in Apple Vision Pro content production, and the quality of depth images ranks first in the industry. From a horizontal perspective, Runway focuses on generative video and does not purely convert 2D to 3D. Depthy and DepthFlow are free but have average accuracy. Its shortcomings are also very clear: it is not a complete 3D modeling tool and cannot change the underlying geometry. It can only make fuss at this layer.
-
The first is pricing friction. When it was originally called LeiaPix, it was basically free. After the name was changed and switched to credit point payment, some old users did not follow, and the free version could not be published. The question of "is it still free or not" has not stopped in the community. The second is that the quality is highly dependent on the original image, and edges such as hair, glass, and thin foregrounds are prone to warping artifacts. The third is cloud dependence, which must be connected to the Internet and has no offline mode. The fourth is that creative control is limited and the geometric structure cannot be changed. Fifth is the lack of large batch processing capabilities, up to 5 at a time. Sixth, the company is a private enterprise, its finances are opaque, and there is no way for the outside world to verify whether it is profitable. The seventh is positioning drift: the official homepage now focuses on the "Immersity Platform" - a set of Spatial AI software plus switchable naked-eye 3D hardware (sold to OEM manufacturers). This is not the same thing as the 2D to 3D conversion tool in the hands of ordinary consumers, and the naming is easy to confuse.
-
It is suitable for photographers, digital artists, XR creators who make spatial computing content, real estate agents who want to use house photos for immersive viewings, musicians who need dynamic covers, and operators who want to run eye-catching social media content. Not suitable for these types of people: people who only convert occasionally and want free export without watermark; people who need complete 3D modeling or character animation; people who want to mass-produce long videos and are sensitive to the cost of credits. If the budget is limited, you can look at DepthFlow (open source and free), Depthy, or the depth toolchains of Runway and ComfyUI as alternatives.
-
Immersity AI can do "turn 2D images into dynamic 3D" accurately and cheaply. It is a rare and convenient tool in the content ecosystem of Vision Pro and Meta Quest. In particular, the dynamic cover of Apple Music is a unique selling point; but if you want to use it seriously, you have to accept credit points for payment, and complex edges and long videos are its weaknesses.
User Reviews
-
Noah.RuizZ—I can’t go back. After seeing the animations I made, the flat photos feel boring. -
HAcha_fi—The watermark in the free version is too discouraging, and it is impossible to publish in 720p. If you want to publish seriously, you still have to buy points. -
JMorris_667—Apple Music’s dynamic cover is so cool, independent musicians save a lot of cover design fees. -
小鱼_16—I tried it on a sunset shot taken during a trip, and it looked like the opening of a blockbuster movie. I posted it on WeChat and was asked several times how I made it. -
dONNA934—For someone who makes a game login interface, it’s a torture if there is no layering in the early stage and it’s difficult to move it in the later stage. Later, I tried Immersity to extract the depth map and then went back to PS to do the replacement. The edge blur is the most natural among these engines. The camera movement intensity is adjusted to about 20%. If it is too large, the picture will be stretched and distorted. The focus is locked in the center to let the background breathe. The exported 4K MP4 can be used directly as a dynamic vertical painting. -
AJackson168—The spatial video exported by Vision Pro was played on Vision Pro. When my friends saw it for the first time, they thought it was shot in real 3D. This is very friendly to people without modeling foundation. -
Karen.Bennett_2022325—Complex backgrounds will occasionally warp, and the probability of hairlines and glass edges being overturned is high, so you have to manually modify the depth map. -
DylanSanchez_2020—I have been using LeiaPix for more than half a year and am an old user who has been following LeiaPix since its free days. To be honest, the depth quality is really good, but credits burn quickly. Exporting a 4K video of more than ten seconds will eat up a lot of credits. Mass production of long videos is still a bit painful. For posting a few pictures on social media every day, the $4.99 picture package is enough. -
LeahArmstrong—Absolutely amazing, just throw the photo away and live. -
Scott_ReedJr—Photographer's perspective: The most valuable thing about it is that it can export the depth map separately. I use Blender to use the displacement map to make real shadows, which is much faster than hand-picking. -
JCox_66—A word of caution if you step on a trap, the resolution of the source image should not be too low. It is recommended to have more than 2 million pixels. It is best to mark the focus manually for night scenes, otherwise the AI will easily recognize distant mountains as nearby ones. In addition, the animation length should be controlled to about 3 seconds for the most comfortable file size. If it is too long, the export will be slow. Overall it is very friendly to novices, just drag it in and adjust the slider. -
BClarkQ—The mobile app is much weaker than the web version, and the key assets are still directed back to the computer. -
枫叶960—E-commerce sellers can use it to dynamically display product images. The conversion rate is indeed better than that of static images, but the cost of credit points must be taken into account. -
Mateo530—After a comparison, Runway's partial generation is not quite pure 2D to 3D, while DepthFlow is free but has average accuracy. Immersity excels in its depth map accuracy and can directly output Vision Pro's MV-HEVC and Meta Quest's SBS. It is almost the only convenient civilian choice for spatial computing content. The disadvantage is that it is not a complete modeling tool and cannot change geometry. -
BlogkHub—yyds, the photos can breathe. -
17jkraa43—New users will receive 100 points, which is enough to play a few games before deciding whether to open a package. -
JoshuaMurray168—Among the 11 sports styles, I use perspective and surround the most. The loop option is very smooth for making short video reels. The preview is free. If you are satisfied, you can save points by switching to 4K. -
jo86ess—From an independent artist’s perspective: Apple Music’s dynamic cover feature is a blow to dimensionality reduction. In the past, making animated covers required a dedicated team to increase the budget. Now, upload the cover and click on the preset to export 3:4 and 1:1 that meet the specifications, and it can also automatically add a blurred background. This is a game changer for people like us who don’t have a design budget, even complex edges will still be distorted occasionally, and the depth map has to be manually modified.