muksoft.Let’s talk
Back to selected work
05 / AI / VIDEO PROCESSING

HyperSwap — The Frame Lab.

Choose the faces. Follow every frame.

Watch demo
MY ROLE

Application development and model integration

PROJECT TYPE

Local video processing application

BUILT WITH

JavaScript · Python · FastAPI · ONNX Runtime · OpenCV · FFmpeg

HyperSwap face selection with a detected face selected and a replacement photo assigned
RECORDED DEMO / 39 SECONDS

See the studio in action.

Download demo
Read the video walkthrough
  1. Upload a video and preview the source footage.
  2. Analyze the video to find and display detected faces.
  3. Select the faces to change and assign a replacement photo to each. Unselected faces stay unchanged.
  4. Start processing and follow actual completed and total frame counts. Preparation and processing are sped up in this recording.
  5. Play the completed result in the studio or download the output video.
01 / THE PROBLEM

Start with the workflow.

Video face swapping involves more than choosing a model. Users need to see which faces were found, decide who changes, assign the right photos, and understand whether a long processing job is moving forward.

02 / THE APPROACH

Connect the pieces.

I built a JavaScript studio and a Python FastAPI processing workflow around existing face-swap models. The interface connects video upload, face analysis, per-face photo assignment, processing progress, and a result player. ONNX Runtime runs model inference, OpenCV supports frame handling, and FFmpeg handles video encoding.

Choose who changes

Review detected faces and assign replacement photos individually, while leaving unselected faces unchanged.

Follow the frames

See actual processed and total frame counts as the application works through the video.

Review the result

Play the completed video in the studio and download the processed MP4.

INSIDE THE APPLICATION

From upload to the final frame.

Five views of the workflow. Open any screenshot at full size in a new tab.

HyperSwap studio with its four-step video workflow
01 / Studio overview — the upload workspace and four-step workflow in the running local application.
Upload and preview the source video
02 / Video upload — preview the source clip before analyzing its faces.
Select detected faces and assign replacement photos
03 / Face assignment — select who changes and assign a replacement photo.
Follow actual processed and total frame counts
04 / Frame progress — completed and total frame counts from the processing job.
Play the completed video or download the MP4
05 / Playback and download — review the completed output and save the MP4.
03 / ENGINEERING DETAIL

The details matter.

Keeping the interface connected to actual processing work. Progress uses completed and total frame counts so users can follow the job through the video, then move directly to playback and download when the result is ready.

An application built around existing face-swap models. My work covers the studio interface, API integration, video workflow, progress reporting, and result handling. The application runs locally and has no public deployment.
UP NEXT / SAAS · WORKSPACE MANAGEMENT

ScopeGrid

✳ HAVE SOMETHING IN MIND?

Let’s build
something useful.

A new product, a smarter workflow, or the next chapter of your app. Tell me what you’re working on.

Start a conversation usamak981@gmail.comPrefer Upwork? Find me there