Real-Time Face Swap AR Filter App
Budget: $1,500 – $3,000 USD
Task Document: https://docs.google.com/document/d/136oLF77gaOnlBTAzUHNO0xN37vBKDijUb5udhX4ozTQ/edit?usp=sharing
The objective of this project is to create an application that enables real-time face swapping using augmented reality (AR) technology, allowing actors to appear in multiple places simultaneously while preserving their likeness, including makeup and glasses. The app will allow users to upload 5 to 10 photos of a face from distinct angles, using a library or custom solution to ensure photo quality and correct angles. These photos will be processed on-device to generate a face model that preserves specific features such as makeup and glasses, and the model will be rendered over a live camera feed in real-time. The swapped face must appear natural and indistinguishable from reality, mimicking all facial movements and expressions (e.g., talking, laughing, yawning, or wearing a nose piercing). The processed video stream will be consumable as a live video feed via an API, mimicking a device camera input, using a gold-standard protocol (e.g., WebRTC) to minimize latency and ensure compatibility, with any deviations communicated to the client. The API must be compatible with Android emulators and devices. Designed for non-technical users, the app must feature a simple, guided interface and include standard camera and video recording controls found in top-tier video recording software. The app is designed with potential future expansion to iOS and web platforms, though no plans exist for multiple simultaneous face swaps.
On request, we will need you to provide the freshly pushed version of your solution via a Dockerhub link. The single repo Docker will have pre-assigned ports for Docker Desktop and include a test Angular app, the main Windows app (and Android app for Milestone 4), and a test Android emulator.
The objective of this project is to create an application that enables real-time face swapping using augmented reality (AR) technology, allowing actors to appear in multiple places simultaneously while preserving their likeness, including makeup and glasses. The app will allow users to upload 5 to 10 photos of a face from distinct angles, using a library or custom solution to ensure photo quality and correct angles. These photos will be processed on-device to generate a face model that preserves specific features such as makeup and glasses, and the model will be rendered over a live camera feed in real-time. The swapped face must appear natural and indistinguishable from reality, mimicking all facial movements and expressions (e.g., talking, laughing, yawning, or wearing a nose piercing). The processed video stream will be consumable as a live video feed via an API, mimicking a device camera input, using a gold-standard protocol (e.g., WebRTC) to minimize latency and ensure compatibility, with any deviations communicated to the client. The API must be compatible with Android emulators and devices. Designed for non-technical users, the app must feature a simple, guided interface and include standard camera and video recording controls found in top-tier video recording software. The app is designed with potential future expansion to iOS and web platforms, though no plans exist for multiple simultaneous face swaps.
On request, we will need you to provide the freshly pushed version of your solution via a Dockerhub link. The single repo Docker will have pre-assigned ports for Docker Desktop and include a test Angular app, the main Windows app (and Android app for Milestone 4), and a test Android emulator.