Faceswap Cards | UXwithmelisa

Face swap Photocards

APP PLATFORM

This project is an exciting innovation that uses AI technology to create an industry-first feature, allowing customers to insert faces from their own photos into specially designed creator led templates, including celebrities, caricatures, and funny characters. The result is a highly personal and humorous greeting card experience that builds on personalisation, boosts engagement, drives growth, and enhances thortful’s proposition for quirky, creative designs whilst supporting independent creators.

Role & Scope

Responsibilities:

Product discovery and problem definition
Defining the UX journey through wireframes, research
Designing end-to-end journeys across web and native
UX and UI design, from concept through launch
Close collaboration with Product, Engineering, Commercial, and Marketing
Post-launch evaluation and iteration planning

Hypothesis - introducing a Face Swap feature will drive business growth for thortful by increasing user engagement, purchase intent, and brand differentiation

Project Problem

Problem Statement

When benchmarking against competitors, other companies are offering more personalised and unique content offering compared to thortful. We see customers expecting more novelty and personalisation in greeting cards, better ways to show their loved ones how much they care. We know uploading photos is common, but swapping faces into designs provides a more playful, surprising experience and is also a first of its kind in the greeting market world. This feature allowed us to differentiate in quite a crowded competitive market.

Customer Problems

Customers are looking for better personalised and unique card content
Increasing prices mean customers are looking for something more bespoke rather than a standard card
Competitors are offering more competitive personalisation, thortful needs to remain competitive to prevent customer churn.

Delivery Milestone 1

MVP on App platform first - test and learn.
To create a BETA launch in 6 weeks to land in tine for Valentine's Day.
Build a cohort of customers for testing group.
Build an A/B experimentation plan alongside marketing to continuously optimise.
If successful launch on Web - Milestone 3.

Outcomes

2% AoV uplift for customers who create a Face swap card.
12K app customers performing nearly 60K face swaps (AVG of 5 swaps per customer).
Resulting in nearly 5K Face Swap cards bought (about 6% of all cards bought on app).
First to market utilising Ai in cards retail

The Final deliverables

Below is a video of how the feature works. If you would like to have a go yourself feature is now live on web HERE.

Key features include:

Original and Selected Face Preview: Users can view both the current and alternative faces side-by-side, helping them easily compare or change the chosen image.

Swap Functionality: A clear, prominent “Swap Face” button allows for quick re-rendering, supporting an effortless trial-and-error process for perfect results.

Edit and Replace Options: The “Change Selected Face” tool enables users to upload or adjust a new photo without restarting the flow.

Guided Simplicity: The design emphasises ease of use, reducing complexity with straightforward icons, minimal text, and a prominent call-to-action.

User Feedback Loop: A “Feedback” button in the header encourages user input, supporting continuous improvement and early issue detection.

Overall, the UX prioritises delight, clarity, and control — ensuring customers can create humorous, high-quality, and personalised cards quickly and confidently. The design balances AI-driven automation with user agency, enhancing trust and reinforcing thortful’s reputation for fun, creative, and accessible innovation.

DISCOVERY - RESEARCH

Problem Defintion

Before diving into the technical feasibility of the Face Swap feature, I wanted to validate the idea from a user and business perspective. To do this, I conducted a series of user interviews with our most loyal customers, alongside discussions with the Marketing and Customer Happiness teams. The goal was to understand what kind of content customers would actually want to create and to align on expectations across key stakeholders.

These early conversations helped establish a shared understanding of the project’s potential and its challenges. While AI-driven experiences are increasingly popular, there was concern that introducing such technology could alienate some users — particularly those wary of privacy issues or unfamiliar with AI tools. It was important to balance innovation with approachability and trust.

Stakeholder questions & concerns:

Through these discussions, a number of broader strategic questions emerged:

Should the Face Swap output remain a simple image swap, or include contextual text to clarify the card’s purpose?
Would message prompts in the personalisation flow help tie the card design and message together more cohesively?
How do we ensure the concept is clear and engaging enough for early adoption, and what role does repeat purchase behaviour play in its success?
What is the right marketing tone to introduce this feature without overwhelming or confusing customers?
How might this project impact thortful’s creator community, who design and sell cards on the platform?
What is the long-term scalability of AI-generated or AI-assisted card content?

There seemed to be two kinds of approach:

1. Let customers type in any prompt and generate a card based on their prompts

2. Create a tool where customers can swap out faces of celebrities with faces of their recipient.

Key insights from research:

1. Customers were concerned about the impact Ai would have on our creators

2. Customers were wary of using Ai incase their information was used inappropriately

3. Customers wanted a bit more guidance when using Ai and therefore open ended prompts wouldn't be helpful

4. Customer Happiness need to engage with customers to reassure them of this new change

Defining the Problem statement

Customers are looking for ways to send better, more personalised content outside of just traditional photocards without also having to use new unknown sites they haven't typically used. Signals from our customers show that lean on funny content to make a deeper more meaning connections are what our customers are looking for from thortful.

DELIVERY

📱Experience Design

Key Journeys

I mapped the full subscription lifecycle, including:

Discovery from PDP, app homepage
First time use of Faceswap, onboarding and tutorials
Completing and uploading a Faceswap
Continuous journey, how can customers easily go back and change their photo without disrupting their flow
Embedding into current site seamlessly

This helped ensure we were designing for customers in mind, the problems that we had identified through research, e.g making them feel confident and safe using Ai. It also helped delivery planning with the eningeering teams to ensure what was being proposed was fully feasible and didn't increase tech debt.

Non subscriber Homepage flow - multi card flow.png

Wire frame flow of MVP Experience design

What were we optimising for?
To help support the business in their understanding of this new proposition. As well as to kick off white boarding with the engineering team to ensure that requirements and scoping could work in parallel to final design deliverables and not have any last minute blockers.
What did we deliberately not do?
There were 3 core behaviours I did not want to include:

Over indexing on this being an Ai feature, we want to encourage customer curiosity rather than cause uncertainty.
Make the journey with too many features for MVP. We wanted customers to be able to create a Faceswap in less than 8 steps
Allow customers to store photos in app storage so that they could come in and out of a session without having to upload a photo each time.

How did research inform my decisions?

Customers wanted to simple and easy to follow UX flows
In this instance, less is more, they didn't want to be overwhelmed with features
They wanted content across every category, this changed our scope slightly as we were intending to launch across Valentine's Day only. But this showed the appetite for more content. How then did I make it easy to find in our catalogue.

MVP for early feedback

The MVP focused on a minimal, streamlined flow designed to test how customers interacted with the Face Swap feature while evaluating whether our third-party AI platform (Segmind) could deliver results within our maximum response time. With a tight deadline to launch ahead of Valentine’s Day, we prioritised building the most efficient version of the tool possible — maintaining essential functionality and high-quality output without unnecessary complexity. This presented a unique challenge, as Face Swap was a completely new concept for customers who might not immediately understand how to use it. Our goal was to create a self-explanatory flow that encouraged exploration and made it easy to retry or adjust results, ensuring users felt in control rather than locked into an outcome they didn’t love.

MVP core features:

Selecting a Face Swap–enabled card template.
Uploading or capturing a photo.
Automatically generating a preview using AI-based face detection and blending.
Providing options to re-swap or change the selected face before finalising the design.

UX flow

User feedback and insight

Core focus

Early feedback was overwhelmingly positive, validating strong novelty and delight value. Users found the experience fun, unique, and highly shareable — confirming that the feature enhanced emotional engagement with the brand.

Interacting with the full journey

Key customer insights included:

Easy to use - most users successfully completed the flow on the first try, praising the simple interface and instant preview.
Greater customisation - some customers mentioned about wanting to swap more than one face, or they would like to put their pets face into the template.
Successful swaps - some feedback did mention that on certain templates, the Ai swap wasn't always super clear. It looked very similar to the photo in the template. Particularly if it was in black and white, it failed to stand out.
Limited Library - For the MVP, we launched with a curated selection of 30 Face Swap–enabled card designs. This smaller library allowed us to focus on testing functionality, performance, and user experience before scaling. However, post-launch feedback revealed that customers often struggled to find a design that suited their recipient or occasion. Many expressed that while the feature itself was fun and engaging, the limited range reduced their likelihood of making a purchase or returning to use it again. This insight highlighted the need to expand the Face Swap catalogue to cater to diverse tastes, relationships, and occasions — a key priority for future iterations to drive repeat engagement and long-term adoption.

Key business insight included:

Higher calls to Ai server - Customers often performed multiple swaps before committing to the purchase of a card. Each swap was approximately 10p and therefore further budgeting needed to be considered to offset any additional costs for the feature.
Consistent traffic - In periods of heavy marketing push, traffic remained quite stable, however once marketing slowed, it wasn't super clear to customers how to navigate to the feature on app.

Data insights

A look back into the last 6 months of FaceSwap has shown it drives its strongest uplift among early-stage and lower-frequency customers.

New & low-frequency buyers show the largest gains:

FaceSwap is particularly effective at converting low-familiarity buyers into repeat users, accelerating them deeper into the lifecycle.

New customers:

+12ppt uplift vs Standard

Single purchasers: +8ppt uplift

Second purchasers: neutral

Conversion Funnels

Total Conversion:

17.1% of users who viewed a Face Swap card completed checkout, indicating strong engagement for an experimental MVP product.

Engagement Flow:

Out of 41,496 users who viewed the product list, 64.9% (26,945) went on to view a specific Face Swap card, and 50.6% (20,989) initiated the Face Swap setup process.
Completion & Preview: Roughly half of those users (50.1%) completed the face swap successfully and viewed the final output, suggesting that the flow and AI performance were generally stable.

Conversion Funnel:

24.3% (10,084) of users proceeded to view the product page post-swap.
20.8% (8,613) added the personalised card to their basket.
17.1% (7,096) completed checkout.

These results demonstrate a healthy level of user curiosity and follow-through, validating the Face Swap feature’s appeal. However, the most notable drop-off occurs between viewing the generated swap and proceeding to the product page (from ~50% to ~24%), indicating a potential opportunity to improve preview satisfaction, loading speed, or visual quality before purchase.