Ultralytics YOLO
Democratizing vision AI
Free
Run YOLO models on-device with real-time camera inference, lens switching, pause-and-share results, and support for detection, segmentation, pose, and more.
Ultralytics YOLO — Vision AI in Your Pocket
Turn your iPhone or iPad into a professional vision-AI tool. Point the camera at the world and see it understood in real time — object detection, instance segmentation, pose estimation, image classification, and oriented bounding box (OBB) detection, all running 100% on-device through the Apple Neural Engine.
Powered by Ultralytics YOLO26 — the latest NMS-free, state-of-the-art real-time detectors from the team behind the open-source YOLO family trusted by millions of developers worldwide.
more It’s actually pretty good, and I do like the fact that you can change the options for the AI. I would highly recommend this, and best of all, its free! But, could you please make it more compatible with the iPhone 8? The bottom of the app is... how do I put it... squished together? And the max number is often just three dots, I would suggest having to scroll down to access the sliders maybe?.... please fix this.
It’s actually pretty good, and I do like the fact that you can change the options for the AI. I would highly recommend this, and best of all, its free! But, could you please make it more compatible with the iPhone 8? The bottom of the app is... how do I put it... squished together? And the max number is often just three dots, I would suggest having to scroll down to access the sliders maybe?.... please fix this.
If you suddenly get a large influx of downloads starting yesterday I promoted the app on a social media channel I own. Great work keep it up been a fan for a while. If you have other stuff you need promo for feel free to reach out to me. Happy to do it.
If you suddenly get a large influx of downloads starting yesterday I promoted the app on a social media channel I own. Great work keep it up been a fan for a while. If you have other stuff you need promo for feel free to reach out to me. Happy to do it.
For the most part the app has a decent UI. The app performs smoothly on the iPhone 11 with no issues I’m glad to see progress being made on the app great job to the team.
For the most part the app has a decent UI. The app performs smoothly on the iPhone 11 with no issues I’m glad to see progress being made on the app great job to the team.
We have deployed Yolov5 into several Jetson NX devices, but this app runs the models much more efficiently than we did.
We have deployed Yolov5 into several Jetson NX devices, but this app runs the models much more efficiently than we did.
• New timing display: see preprocessing, inference, and postprocessing per frame under the FPS readout
• Sharper semantic segmentation: full-resolution masks with crisp class boundaries, rendered faster
• Smoother camera switching: the transition now covers the full interface
• Layout fixes: model selector sized correctly in landscape, no placeholder text or layout shift at launch
8.9.6 Jun 11
Version 8.9.4 adds camera torch control for brighter real-time inference and brings the latest YOLOView SDK updates into the app.
• Added a camera torch toggle for live YOLO inference
• Added YOLOView support for custom overlay rendering
• Added live instance-mask streaming for segmentation workflows
• Improved segment mask accuracy around detection boxes
• Fixed storyboard loading for YOLOView-based integrations
• Includes SDK, documentation, and reliability updates for the shared UltralyticsYOLO package
8.9.4 Jun 7
- Significant performance improvements, including faster per-frame postprocessing, Apple Neural Engine inference, and 720p camera capture.
- Added an in-app info drawer with quick links to Ultralytics Docs, YOLO26 models, GitHub, and Licensing.
- Improved model switching reliability with cancel support for remote downloads on slow networks.
- Improved loading overlay behavior during device rotation.
- Refined bottom toolbar spacing on modern iPhone displays.
8.8.8 Jun 1
- Added support for YOLO semantic segmentation models.
- Improved on-device vision workflows for dense pixel-level scene understanding.
- Includes stability and compatibility refinements for the latest YOLO model support.
8.8.7 May 22
- Dedicated external display now works in both portrait and landscape — no need to rotate your iPhone
- Camera orientation on the external screen follows your device, not the monitor
- Detection boxes, segmentation masks, pose keypoints, and OBBs are now correctly placed on external displays
- New "Dedicated External Display" toggle in iOS Settings (off by default), applied without restarting the app
- iPad landscape support
8.8.6 May 12
- Fixed YOLO camera preprocessing to preserve image aspect ratio instead of stretching frames.
- Improved portrait and landscape inference orientation so objects are presented upright to the model.
- Added better support for square, portrait-only, and landscape-only Core ML model input shapes.
- Updated YOLO26 model export guidance for iOS and Core ML.
- Improved App Store links and documentation for international users.
8.8.5 May 9
- Added an iOS-style camera lens selector for switching between available rear camera lenses.
- Improved camera switching so the rear camera returns to the expected 1x view.
- Fixed share behavior when the camera is paused, including sharing the paused frame with detection boxes.
- Improved share capture reliability while the camera is running.
- Added a quick info button linking to Ultralytics.
- Improved camera controls, zoom selector sync, and overall camera UI polish.
- General reliability and performance improvements.
8.8.4 May 1
A major under-the-hood refresh focused on speed, accuracy, and polish.
• Powered by the latest Ultralytics YOLO26 models — NMS-free, faster, and more accurate on small and distant objects
• Streamlined inference pipeline with unified rendering across detect, segment, pose, OBB, and classify
• More responsive thresholds: sliders update instantly and stay in sync with the running model
• Faster startup and clearer error messages when loading custom models
• Fixed an OBB overlay label that displayed a stray percent sign
• Reduced memory use and CPU overhead — ~15% smaller SDK footprint
• Dozens of refinements to logging, threading, and error handling for production-grade reliability
Feedback and bug reports welcome at github.com/ultralytics/yolo-ios-app — thanks for using Ultralytics YOLO.
8.8.2 Apr 18
Now powered by the latest Ultralytics YOLO26 models for faster, more accurate on-device detection.
- Improved real-time tracking stability and responsiveness
- Better recognition of small and distant objects
- General performance and reliability improvements
8.8.1 Feb 8
Enhanced UI with improved model selection display, responsive font sizing, better spacing, and optimized layout for all device sizes.
8.7.27 10/05/2025
Fixed the bug in external display mode.
8.7.18 09/23/2025
- Added External Display Support for bigger, clearer, and more powerful viewing.
- Performance improvements and minor bug fixes for a smoother experience.
8.7.17 09/22/2025
- Faster autodownload of CoreML models
- Optimized inference code
- Full support for all 5 YOLO tasks
8.7.13 08/28/2025
- Faster autodownload of CoreML models
- Optimized inference code
- Full support for all 5 YOLO tasks
8.7.12 08/04/2025
Fixed a bug that caused the UI to break.
8.5.9 04/14/2025
Fixed a bug that caused the camera image to rotate 90 degrees when launched when the device orientation was landscape.
Fixed a bug that caused object count to not be reflected in landscape mode.
Yolo detection models are now sorted by size.
8.5.8 03/08/2025
Fixed a bug that caused a crash on startup on some OS.
Added classify, segment, pose, obb tasks.
8.5.2 03/05/2025
Bug fixes and performance improvements in landscape mode
8.4.0 11/13/2024
Performance has improved with the addition of new models.
8.3.0 11/06/2024
Version 8.2.0 brings exciting updates to enhance your experience:
- NEW YOLOv8 models: We've fine-tuned our YOLO models for even better object recognition.
- Improved User Interface: Enjoy a sleeker, more intuitive design that makes navigation smoother than ever.
- Performance Boosts: Experience faster, more responsive object detection and image classification.
- Bug Fixes: We've squashed some bugs for a smoother app performance.
We continuously improve to bring you the best object detection experience on mobile. Update now and keep discovering the world in exciting new ways!
8.2.0 02/12/2024
Version 8.1.0 brings exciting updates to enhance your experience:
- NEW YOLOv8 models: We've fine-tuned our YOLO models for even better object recognition.
- Improved User Interface: Enjoy a sleeker, more intuitive design that makes navigation smoother than ever.
- Performance Boosts: Experience faster, more responsive object detection and image classification.
- Bug Fixes: We've squashed some bugs for a smoother app performance.
We continuously improve to bring you the best object detection experience on mobile. Update now and keep discovering the world in exciting new ways!
8.1.0 02/11/2024
- New YOLOv5-v6.0 models including YOLOv5n Nano
- Bug fixes and performance improvements
8.0.0 10/15/2021
-New YOLOv5 release 5.0 models with nn.SiLU() activations
-Realtime IoU threshold and confidence threshold updates using a slider bar
-Bug fixes and performance improvements
7.9 09/12/2021
-YOLOv5.3 models with nn.Hardswish() activations
-Realtime IoU threshold and confidence threshold updates using a slider bar
-Landscape mode and upside down mode inference
-Bug fixes and performance improvements
7.8 10/28/2020
-Updated to use the latest YOLOv5.3 models with hardswish activations
-Added capability for realtime IoU threshold and confidence threshold updates using a slider bar
-Added capability for landscape mode and upside down mode inference
-Bug fixes and performance improvements
7.7 09/20/2020
• New timing display: see preprocessing, inference, and postprocessing per frame under the FPS readout
• Sharper semantic segmentation: full-resolution masks with crisp class boundaries, rendered faster
• Smoother camera switching: the transition now covers the full interface
• Layout fixes: model selector sized correctly in landscape, no placeholder text or layout shift at launch
more Version 8.9.6 Jun 11
Data Not Collected The developer does not collect any data from this app.
