The latest articles on DEV Community by FacePika (@facepika). https://dev.to/facepika https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4190287%2F928eb167-d3de-4073-9985-9eaa4b46fb2d.jpg https://dev.to/facepika en FacePika Sat, 17 Oct 2026 10:55:47 +0000 https://dev.to/facepika/how-we-built-an-on-device-face-swap-app-for-android-with-open-weight-ml-models-no-cloud-uploads-132m https://dev.to/facepika/how-we-built-an-on-device-face-swap-app-for-android-with-open-weight-ml-models-no-cloud-uploads-132m <p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwr2tjf16mpa3gzvi1oa2.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fwr2tjf16mpa3gzvi1oa2.png" alt=" " width="800" height="268"></a></p> <p>When we started building FacePika, we had one hard constraint: users' photos and videos should never leave their phone.</p> <p>Most face swap tools today upload your media to a remote server, run the model in the cloud, and send the result back — that works, but it creates obvious privacy tradeoffs. We wanted to build something that worked fully offline, on mid-range Android phones, without needing a data center behind it.</p> <p> <iframe src="https://www.youtube.com/embed/5O3PiZUDWQA" width="710" height="399"> </iframe> </p> <h2> What we learned deploying open-weight models on-device </h2> <p>The biggest challenge wasn't running the model itself — it was making it fast enough and small enough to feel usable on a regular phone:</p> <ul> <li>We started with an open-source face swap model, and quantized it down from FP32 to INT8, cutting the model size by 75% without noticeable quality loss.</li> <li>We used Android's NNAPI to offload inference to the device's GPU/NPU, which cut processing time for a single photo swap from 12 seconds to under 2 seconds on a mid-range 2024 phone.</li> <li>Face alignment was trickier than we expected: prepping the face landmarks before feeding them to the model ended up being 30% of our total processing time, and we ended up writing our own lightweight landmark detector to replace a heavier library.</li> </ul> <h2> Why on-device matters </h2> <p>The best part of this approach is there's no upload step at all: a user picks a photo, the swap happens entirely in app, and the output never touches a remote server. For a tool that processes people's faces, that privacy guarantee isn't an afterthought — it's the whole point.</p> <p>We're still refining the model and adding more template support, and we're planning to write more about our mobile ML optimization process here as we go.</p> <p>The app we built is called FacePika — you can check it out here: <a href="https://facepika.app/" rel="noopener noreferrer">facepika.app</a></p> ai android machinelearning mobile