# Performance benchmarks

> Measured frame rates, face detection times, and SDK size for Nosmai Effects on real Android phones and iPhones.

> For AI agents: the complete documentation index is at https://nosmai.com/llms/effects.txt

Product: Nosmai Effects
SDK version: Android 3.0.7 · iOS 3.0.7 · Flutter 2.0.2 · React Native 2.0.1 · Web 0.2.0
Group: reference
Source: https://nosmai.com/docs/effects/performance-benchmarks/



Nosmai Effects runs beauty, AR effects, virtual backgrounds, recording, and live
streaming in real time on everyday Android and iOS devices. Every result on this
page was measured on a physical device running Nosmai Effects SDK 3.0.7.

## At a glance

| Result | Workload |
| --- | --- |
| **30 FPS** | Real-time beauty preview on mid-range and high-end Android phones and iPhones |
| **28.6 to 29.9 FPS** | Recorded 720p video with beauty enabled |
| **29.4 to 30.0 FPS** | Stream-ready video output with beauty enabled |

## Effect frame rates

The camera preview frame rate with each feature running on its own. **FPS**
(frames per second) is the number of processed camera images shown each
second. Nosmai Effects targets 30 FPS, the standard rate for mobile camera
apps.

### Android

| Device | Chip | OS | Beauty FPS | AR effect FPS | Background FPS |
| --- | --- | --- | --- | --- | --- |
| Samsung Galaxy S22 Ultra | Snapdragon 8 Gen 1 | Android 14 | 30.0 | 29.0 | 29.0 |
| Google Pixel 7 | Google Tensor G2 | Android 17 | 30.1 | 30.0 | 30.0 |
| Google Pixel 4 | Snapdragon 855 | Android 13 | 30.0 | 29.0 | 28.0 |
| Motorola Edge 2024 | Snapdragon 7s Gen 2 | Android 14 | 30.0 | 28.0 | 28.0 |
| Google Pixel 5 | Snapdragon 765G | Android 14 | 30.0 | 29.0 | 30.0 |
| OnePlus Nord N30 | Snapdragon 695 | Android 14 | 30.1 | 29.5 | 28.4 |
| Redmi Note 10 | Snapdragon 678 | Android 13 | 29.0 | 30.0 | 27.0 |
| OnePlus Nord N10 5G | Snapdragon 690 | Android 11 | 29.9 | 29.0 | 30.9 |
| OnePlus Nord N200 5G | Snapdragon 480 | Android 12 | 30.0 | 30.0 | 30.1 |

### iOS

| Device | Chip | OS | Beauty FPS | AR effect FPS | Background FPS |
| --- | --- | --- | --- | --- | --- |
| iPhone 15 Plus | A16 Bionic | iOS 26.2 | 30.0 | 30.0 | 30.0 |
| iPhone 12 | A14 Bionic | iOS 26.6.2 | 30.0 | 29.0 | 28.0 |
| iPhone 11 | A13 Bionic | iOS 26.6.2 | 30.0 | 29.0 | 28.0 |
| iPhone SE (2020) | A13 Bionic | iOS 26.6.2 | 30.0 | 28.0 | 28.0 |
| iPhone XS Max | A12 Bionic | iOS 18.7.10 | 30.0 | 29.0 | 30.0 |
| iPhone XR | A12 Bionic | iOS 18.6.2 | 30.0 | 27.0 | 30.0 |
| iPhone X | A11 Bionic | iOS 16.7.16 | 29.0 | 25.0 | 30.0 |

## Beauty, recording, and streaming

These results show what your users see and what your app saves or sends while
a beauty look is active. Each device runs the same look: skin smoothing, skin
whitening, lipstick, and eyebrows. Video output is 720 × 1280 at a 30 FPS
target.

- **Preview FPS:** the live camera preview with the beauty look applied.
- **Recorded FPS:** the frame rate of the saved video file.
- **Streaming FPS:** the frame rate of the encoded video your app sends to a
  streaming service.
- **Recording + streaming FPS:** recording and streaming at the same time.

### Android

| Device | Chip | OS | Preview FPS | Recorded FPS | Streaming FPS | Recording + streaming FPS |
| --- | --- | --- | --- | --- | --- | --- |
| Samsung Galaxy S22 Ultra | Snapdragon 8 Gen 1 | Android 14 | 30.0 | 29.7 | 29.6 | 29.8 |
| Google Pixel 7 | Google Tensor G2 | Android 17 | 30.1 | 29.5 | 29.5 | 29.4 |
| Google Pixel 4 | Snapdragon 855 | Android 13 | 30.0 | 29.6 | 29.7 | 29.5 |
| Motorola Edge 2024 | Snapdragon 7s Gen 2 | Android 14 | 30.0 | 29.9 | 30.0 | 30.0 |
| Google Pixel 5 | Snapdragon 765G | Android 14 | 30.0 | 28.6 | 29.4 | 29.5 |
| OnePlus Nord N30 | Snapdragon 695 | Android 14 | 30.1 | 29.6 | 29.6 | 29.7 |
| Redmi Note 10 | Snapdragon 678 | Android 13 | 29.5 | 28.9 | 29.4 | 28.8 |
| OnePlus Nord N10 5G | Snapdragon 690 | Android 11 | 29.9 | 29.3 | 29.5 | 29.2 |
| OnePlus Nord N200 5G | Snapdragon 480 | Android 12 | 29.6 | 28.8 | 29.4 | 28.6 |

### iOS

| Device | Chip | OS | Preview FPS | Recorded FPS | Streaming FPS | Recording + streaming FPS |
| --- | --- | --- | --- | --- | --- | --- |
| iPhone 15 Plus | A16 Bionic | iOS 26.2 | 30.0 | 29.9 | 30.0 | 29.9 |
| iPhone 12 | A14 Bionic | iOS 26.6.2 | 30.0 | 29.7 | 29.9 | 29.7 |
| iPhone 11 | A13 Bionic | iOS 26.6.2 | 30.0 | 29.7 | 29.9 | 29.7 |
| iPhone SE (2020) | A13 Bionic | iOS 26.6.2 | 30.0 | 29.7 | 29.9 | 29.7 |
| iPhone XS Max | A12 Bionic | iOS 18.7.10 | 30.0 | 29.2 | 30.0 | 29.2 |
| iPhone XR | A12 Bionic | iOS 18.6.2 | 30.0 | 29.4 | 29.9 | 29.3 |
| iPhone X | A11 Bionic | iOS 16.7.16 | 29.5 | 29.0 | 29.6 | 28.9 |

## Face detection time

The time Nosmai Effects takes to find a face in a single camera frame, in
milliseconds (ms). Lower is better. At 30 FPS, each frame has about 33 ms in
total, so a fast detection leaves most of that time for effects and rendering.

### Android

| Device | Chip | Average (ms) | 95th percentile (ms) |
| --- | --- | --- | --- |
| Samsung Galaxy S22 Ultra | Snapdragon 8 Gen 1 | 3.78 | 4.31 |
| Google Pixel 7 | Google Tensor G2 | 5.25 | 6.58 |
| Google Pixel 4 | Snapdragon 855 | 3.95 | 4.80 |
| Motorola Edge 2024 | Snapdragon 7s Gen 2 | 2.91 | 3.18 |
| Google Pixel 5 | Snapdragon 765G | 5.20 | 10.44 |
| OnePlus Nord N30 | Snapdragon 695 | 2.66 | 3.60 |
| Redmi Note 10 | Snapdragon 678 | 6.10 | 8.20 |
| OnePlus Nord N10 5G | Snapdragon 690 | 5.40 | 7.10 |
| OnePlus Nord N200 5G | Snapdragon 480 | 7.20 | 9.60 |

### iOS

| Device | Chip | Average (ms) | 95th percentile (ms) |
| --- | --- | --- | --- |
| iPhone 15 Plus | A16 Bionic | 2.10 | 2.60 |
| iPhone 12 | A14 Bionic | 2.70 | 3.30 |
| iPhone 11 | A13 Bionic | 3.10 | 3.80 |
| iPhone SE (2020) | A13 Bionic | 3.20 | 3.90 |
| iPhone XS Max | A12 Bionic | 3.60 | 4.50 |
| iPhone XR | A12 Bionic | 3.70 | 4.60 |
| iPhone X | A11 Bionic | 4.40 | 5.60 |

The **95th percentile** means 95% of frames were processed at or below this
time. It shows how consistent detection stays, not only how fast it is on
average.

> **Note**
>
> **Older Android devices**
>
> Some older and entry-level Android devices, especially 32-bit (ARMv7) models
> from brands such as Oppo, Tecno, Infinix, Redmi, and Samsung, currently run
> Nosmai Effects at lower frame rates. All features are supported on these
> devices, and performance improvements for them are in active development.

## SDK size

The size Nosmai Effects SDK 3.0.7 adds to your app. The App Store and Google
Play deliver only the build for each user's device architecture, so a user
downloads one row of this table, not all of them.

| Platform | Download size | Installed size |
| --- | --- | --- |
| iOS (arm64) | 27.2 MB | 42.6 MB |
| Android (arm64-v8a) | 31.4 MB | 58.5 MB |
| Android (armeabi-v7a) | 22.7 MB | 34.0 MB |

**Download size** is the compressed size a user downloads. **Installed size**
is the space used on the device. Final sizes can vary slightly with each app
store's compression. For the published file sizes of each release, see
[Limits and performance](/docs/effects/limits-and-performance/).

## How we tested

| Item | Setting |
| --- | --- |
| SDK version | Nosmai Effects 3.0.7 |
| Test period | October 2026 |
| Camera | Front camera, live preview |
| Video output | 720 × 1280, H.264, 30 FPS target, 4 Mbps |
| Beauty look | Skin smoothing, skin whitening, lipstick, and eyebrows |
| Beauty, recording, and streaming | 5-second warm-up followed by a 20-second measurement for each mode |
| Face detection | 10-second warm-up followed by a 30-second measurement |
| Effect frame rates | 10-minute session for each feature |

Performance varies with the device, lighting, face visibility, device
temperature, resolution, and the effects you enable. These results are a
reference, not a guarantee for every device or effect combination. Measure
Nosmai Effects with your final effects and settings on the devices your users
own.

## Try it on your devices

Start a 30-day free trial with every feature included and measure Nosmai
Effects in your own app. [Contact us](https://nosmai.com) to discuss your use
case or schedule a demo.
