Contact person: Business Client

Phone:Show

Email:Show

Location: Lisbon, Portugal

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I want a lightweight microsite that lets any visitor input a frontal human face photo and a frontal animal face to obtain the transformation video between the two.

The first release has to support dog and cat outputs; building the code in a way that makes “other” animals easy to add later will be a plus.

Flow
A user lands on the page, taps “Upload”, selects the pictures directly from their device, and within seconds receives a transformation video between the two that can be downloaded locally. No URL uploads are needed right now. The uploader must accept JPEG and PNG only and warn gracefully if another format is chosen.

Usage is free. We do not store any identification or images. During the process, the user sees a 15/20-second advertisement.

Core requirements
• Front-end: snappy single page, responsive, clean UI/UX that guides the user through upload → processing → download.
• Back-end: the face-to-face swap itself. Whether you prefer a GAN, pre-trained diffusion model, or classical morphing, the main thing is a convincing dog/cat rendering that keeps the original pose and lighting believable. Open-source stacks such as TensorFlow, PyTorch or ONNX are fine as long as the finished service runs on a modest cloud instance.
• Output: downloadable file (same format as upload) with resolution not degraded below 1024 px on the longest side.
• Privacy: discard uploads after processing; no storage beyond the current session.
• Extensibility: clean code and model pipeline so a new animal can be added with minimal retraining or asset swap.

Deliverables
1. Deployed microsite (sub-domain or lightweight hosting) ready for public use.
2. Full source code with install notes and model weights or links.
3. Quick README covering how to plug in an additional animal template later.

I will test with a set of sample faces and expect the average turnaround per image to stay under 20 s on commodity hardware. Once these checkpoints are met and the site is live, the project is complete." (client-provided description)