MagiScan

What is MagiScan?

MagiScan is a mobile-based 3D scanning and AI modeling application developed by Magiscan Inc.; it enables the conversion of real-world objects, photos taken from various angles, individual images, or textual descriptions into 3D models. It is not a platform designed for users to train their own machine learning models.

The products are available for iPhone, iPad, and Android devices; they offer desktop model management, import capabilities for Unity and NVIDIA Omniverse, integration with 3D printing software, as well as 3D/AR display options for e-commerce use. The camera and LiDAR capabilities of the devices, along with their respective packages, affect the available modes.

Main functions

• 360 Scan: The app guides users to take photos from all angles around a stationary object, making it suitable for those using smartphone photogrammetry for the first time.
• Freeform scanning: The user controls the shooting angle and path, making it suitable for objects with complex shapes or those that require additional shots of their bottom or grooves.
• Hybrid scanning: Combines depth and photographic information to improve coverage of geometry and texture; the actual results depend on the device’s sensors.
• Photo album upload: It is possible to import photos taken from various angles using a phone’s camera, a drone, or a professional camera; these photos can be submitted for reconstruction when there is an internet connection.
• Converting a single image to 3D: MagiSnap allows you to take a photo directly, while One Image to 3D lets you select an image from the album; it’s suitable for quickly creating conceptual models.
• Text to 3D: The latest version allows for the creation of models from textual descriptions, without the need to scan real objects first.
• 3D Viewer: Offers perspective and orthographic views, as well as options for adjusting the background, brightness, and lighting; it also allows for direct opening of models from completion notifications.
• Automatic lighting compensation: When the ambient light is insufficient, Android states that the app can automatically activate the phone’s flash to assist with photography.

LiDAR, room, and point cloud modes

• Room Plan uses AR and LiDAR to quickly create detailed floor plans of rooms, with support for both metric and imperial units.
• Room Plan requires an Apple device equipped with LiDAR; the app’s instructions are available starting with the iPhone 12 Pro series.
• A Point Cloud can scan areas and environments within a range of up to five meters, offering automatic or manual data collection, as well as the ability to adjust the size of the points and their display.
• The point cloud mode is processed on the device side, and files in PLY, XYZ, PTS, and PCD formats can be exported; some features remain marked as Premium or available only during the trial period.
• LiDAR is suitable for capturing spatial outlines and point clouds, but it does not guarantee the precision required for surveying, manufacturing inspection, or medical devices.

Scan input and model output

Export format

Scan tutorial

1. Install the official app on an iPhone, iPad, or Android device, create an account and grant permission for access to the camera; grant access to the photo album only when it’s time to import photos.
2. First, select a stationary object of moderate size, with distinct surface texture and no reflective properties; place it in a location where the lighting is even and the background is simple.
3. Beginners should choose the 360-degree mode: move slowly around the object, maintaining a relatively constant distance, and cover the top, sides, as well as any shaded areas.
4. After the collection is completed, it is submitted for processing, and the estimated time in the queue can be viewed. Free tasks may have to wait for a longer time, while Premium users generally receive priority handling.
5. After the model is prepared, it can be rotated and enlarged in the viewer to check for holes, double images, stretching, floating fragments, misaligned textures, and a missing bottom portion.
6. Choose the format based on its intended use; before 3D printing, check the units, completeness, and wall thickness, and reduce the number of polygons and texture size before using it in games or web pages.
7. Export to local devices, MagiScan Studio, Unity, Omniverse, or slicing software, and carry out cleaning and dimension calibration in the target tool.

Improve scanning quality

• Keep the object stationary and move the camera slowly; there should be sufficient overlap between consecutive photos, and the back and top of the object must not be missed.
• Use uniform diffused light, avoiding strong direct light, deep shadows, and fluctuations in exposure; complex backgrounds can interfere with feature matching.
• Transparent, mirror-like, black, or smooth surfaces are prone to failure; when necessary, use safely removable matte treatments or temporary texture markings.
• Start with a free trial before deciding whether to pay; mobile scanning is suitable for quick digitization, but it cannot replace calibrated industrial measurement equipment.
• When dealing with human bodies, medical devices, cultural relics, and precision parts, it is necessary to use reliable dimensional references as well as professional verification methods.

Personal version price

• The amounts mentioned above are for the US App Store; prices vary in other countries, on Google Play, as well as due to taxes, promotions, and account-based pricing.
• The store states that unlimited use is available after subscription, while the Premium version mentions an increase of 10 scans per day; there is a contradiction in the descriptions of the benefits, and it should be the payment page and the account quota that determine the actual rights.
• The additional scanning quota available each day is accumulated and retained, but the specific rules regarding expiration, suspension of the subscription, and resumption of purchases should be checked in the account settings.
• Subscriptions are managed and renewed by the respective app stores, and cancellation and refunds must follow Apple’s or Google’s procedures.

Corporate 3D and AR solutions

The company’s pricing page does not clearly indicate the settlement period for the aforementioned amounts; the terms mention advance payment and a validity period of twelve months. Before making a purchase, it is necessary to request a formal quote in order to confirm the number of models, the number of views, any excess fees, the renewal date, and the service level.

Platforms and integration

API, GitHub, and open-source status

• MagiScan’s core 3D and AI engines are explicitly designated as proprietary technologies, while its mobile applications and model services are not open-source products.
• The official GitHub organization has made available the MagiScan for Unity package, which includes an editor window, functionality for connecting to accounts via QR codes, a read-only task API client, as well as examples for downloading and running GLB files.
• The Unity repository is publicly readable, but its license is explicitly designated as proprietary; it allows free installation and use for connecting to the MagiScan service only, and it should not be mistaken for an open-source SDK.
• The integration token for Unity packages can be revoked from the mobile device; by default, the editor stores the token in each user’s settings, but projects that are run should employ a more secure method for storing tokens.
• There are no confirmed general cloud APIs available for third parties, nor any information regarding public pricing, rate limits, or service levels.

Model ownership and commercial risks

• The terms state that the user owns the 3D models generated by MagiScan.
• The same clause also allows MagiScan to use these models on its own, including for machine learning training and the development of generative AI products.
• The platform also retains the right to sell the model separately or to provide it to third parties as part of a larger dataset, without having to pay royalties or any other compensation to the users.
• Therefore, it is not appropriate to scan business secrets, unreleased products, customer samples, copyrighted works, human bodies, or sensitive areas without first conducting an assessment.
• Users still need to have the rights to the object being scanned, to the photos, and to the prompts used for generation; ownership of the model does not eliminate the copyright, patents, trademarks, privacy rights, and other restrictions related to the original object.

Privacy and data security

• After authorization, the app will have access to the camera and photo album; photos can be uploaded to MagiScan or the service provider’s servers, or they can be stored on the device itself.
• The privacy policy permits the collection of IP addresses, device identifiers, system and browser information, usage records, and diagnostic data, which are used for services, analysis, marketing, and product improvement.
• The privacy policy states that personal data should be retained only for as long as it is necessary for the intended purpose; it does not specify any fixed deadline for deleting scanned photos, point clouds, models, and backups.
• Users can delete certain information from their accounts or contact support to request access, correction, or deletion; however, legal obligations or other legitimate reasons may require that certain data be retained.
• Data may be processed across borders; the platform pledges to implement reasonable control measures, but it cannot guarantee absolute security for data transmission or electronic storage.
• The information provided by the stores varies: the Google Play page states that no data is collected, but it is possible that location and device identifiers are shared and the data cannot be deleted; the Apple page lists location data, user-generated content, identifiers, as well as usage and diagnostic data. It is necessary to make a careful judgment based on the actual platform and its permission settings.

Capacity boundaries

• Objects that are reflective, transparent, black, elongated, have repetitive textures, or lack surface features are prone to developing holes, deformation, and misaligned textures.
• The free queue can be slow, and even priority processing for paid users does not guarantee that each task will be completed by the expected time.
• Image and text generation can infer the invisible parts, but it is not suitable for restoring the actual size, mechanical structure, or details of an evidential nature.
• Mobile phone scanning models usually require further cleanup, face removal, hole fixing, scaling, and material optimization in Blender, Unity, or mesh processing tools.
• Room dimensions and point clouds cannot replace the calibration equipment and specialized procedures required for building surveys, medical scanning, quality inspection, or reverse engineering.

Summary

MagiScan focuses on converting images taken with a smartphone, data from LiDAR, individual photos, and text inputs into exportable 3D assets, while also enabling integration with Unity, Omniverse, 3D printing, and e-commerce AR solutions. Individual users can first use the free scanning service to determine whether a particular object is suitable for processing, and then choose a subscription plan or scanning package based on their priorities, quotas, and desired format.

The most significant procurement risk lies not in the scanning accuracy, but rather in the terms related to the model data. The platform retains the rights to train the models, to use them for commercial purposes, and to sell the datasets; when dealing with proprietary products, customer data, or information related to human bodies, it is necessary to obtain authorization before uploading such data and to reassess whether it is appropriate to use cloud services in such cases.