---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: image-to-video

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/fomm/web-assets/model_demo.png)

# First-Order-Motion-Model: Optimized for Qualcomm Devices

FOMM is a machine learning model that animates a still image to mirror the movements from a target video.

This is based on the implementation of First-Order-Motion-Model found [here](https://github.com/AliaksandrSiarohin/first-order-model/tree/master).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.57.1/src/qai_hub_models/models/fomm) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.45, ONNX Runtime 1.25.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/fomm/releases/v0.57.1/fomm-onnx-float.zip)

For more device-specific assets and performance metrics, visit **[First-Order-Motion-Model on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/fomm)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/v0.57.1/src/qai_hub_models/models/fomm) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [First-Order-Motion-Model on GitHub](https://github.com/qualcomm/ai-hub-models/blob/v0.57.1/src/qai_hub_models/models/fomm) for usage instructions.

## Model Details

**Model Type:** Model_use_case.video_generation

**Model Stats:**
- Model checkpoint: vox-256
- Input resolution: 256x256
- Model size (detector) (float): 54.2 MB
- Model size (generator) (float): 174 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| detector | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.25 ms | 0 - 27 MB | NPU
| detector | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 7.354 ms | 1 - 26 MB | NPU
| detector | ONNX | float | Qualcomm® QCS8550 (Proxy) | 4.394 ms | 0 - 32 MB | NPU
| detector | ONNX | float | Qualcomm® QCS8450 | 7.354 ms | 1 - 26 MB | NPU
| detector | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.743 ms | 1 - 21 MB | NPU
| detector | ONNX | float | Qualcomm® QCS9075 | 5.804 ms | 1 - 46 MB | NPU
| detector | ONNX | float | Snapdragon® 8 Elite Mobile | 2.914 ms | 0 - 22 MB | NPU
| detector | ONNX | float | Qualcomm® QCS8750 | 2.914 ms | 0 - 22 MB | NPU
| detector | TFLITE | float | Qualcomm® SA8775P | 11.63 ms | 1 - 19 MB | GPU
| detector | TFLITE | float | Qualcomm® SA8650P | 11.63 ms | 1 - 19 MB | GPU
| detector | TFLITE | float | Qualcomm® SA8255P | 11.63 ms | 1 - 19 MB | GPU
| generator | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 16.359 ms | 18 - 199 MB | NPU
| generator | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 37.095 ms | 16 - 194 MB | NPU
| generator | ONNX | float | Qualcomm® QCS8550 (Proxy) | 21.762 ms | 18 - 21 MB | NPU
| generator | ONNX | float | Qualcomm® QCS8450 | 37.095 ms | 16 - 194 MB | NPU
| generator | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 10.96 ms | 16 - 182 MB | NPU
| generator | ONNX | float | Qualcomm® QCS9075 | 34.927 ms | 18 - 63 MB | NPU
| generator | ONNX | float | Snapdragon® 8 Elite Mobile | 13.285 ms | 16 - 178 MB | NPU
| generator | ONNX | float | Qualcomm® QCS8750 | 13.285 ms | 16 - 178 MB | NPU
| generator | TFLITE | float | Qualcomm® SA8775P | 819.964 ms | 21 - 37 MB | CPU
| generator | TFLITE | float | Qualcomm® SA8650P | 819.964 ms | 21 - 37 MB | CPU
| generator | TFLITE | float | Qualcomm® SA8255P | 819.964 ms | 21 - 37 MB | CPU

## License
* The license for the original implementation of First-Order-Motion-Model can be found
  [here](https://github.com/AliaksandrSiarohin/first-order-model/blob/master/LICENSE.md).

## References
* [First Order Motion Model for Image Animation](https://arxiv.org/abs/2003.00196)
* [Source Model Implementation](https://github.com/AliaksandrSiarohin/first-order-model/tree/master)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).