- Home
- 2025 projects
- GovStock
GovStock
10 February 2025
Find the perfect photo faster for your government communications—with authentic Singaporean models and scenes

Problem
Government communication officers often create marketing collaterals and website content that has images. Here’s the issue though—it’s really hard for them to find images that are high quality, fit the content, and most importantly, reflect the multiracial makeup of Singapore.
It takes at least half an hour per image to source
…and that’s considered fast. More often than not, it can actually take much longer to source for an image that will work.
The first hurdle is deciding what kind of imagery suits your content best. Then you need to find the image by going through stock photo libraries. These stock libraries are based overseas, so finding images with authentic Singaporean faces, clothing, and settings can be a challenge—not to mention representation of racial and age diversity and other key concerns for government officers. The pain doesn’t end there—once your image gets approved, the process to procure a license to use the image can take up to a few weeks.
And of course, it all costs taxpayer money.
Some officers turn to DIY (Do-It-Yourself)
In many of our interviews, we’ve heard of comms officers having to do their own photoshoots in their offices or “hacking out” a banner by using PowerPoint. This not only results in poorer quality images on government communications but also is a time sink for many officers.
Poor quality images erode trust
You might think, "what’s the problem though, isn’t it just a picture?"
Not quite. Poor-quality images can erode trust in government communications. Authentic images also provide a human face to government services and engage citizens in the messages agencies care about communicating.
Ultimately, our goal is to give government better tools to build trust and engage members of the public in their communications.
Solution
To tackle this problem, we built GovStock.
We aimed to achieve the following:
Slash sourcing time: From 30+ mins to >10 mins per image
Reduce approval time: By putting more relevant and quality images at officers’ fingertips
Improve trust: By ensuring images in government comms are authentically Singaporean and high quality
A library of authentic photos with Singaporean faces and scenes

Images can be filtered by categories relevant to government users
The first thing you see on GovStock is a stock library of authentic photos featuring Singaporean people and scenes.
Comms officers can search for photos by entering a search term or clicking a category. We hand-picked categories that appear frequently on government comms, such as Sustainability & Environment or Public Health & Wellness.
Localised tags
Images are tagged with tags that are significant to the local audience. For example, instead of calling a residential building “apartment”, we label it as “HDB”. Tags such as “park connector”, “void deck”, and “hawker center” help public officers find the perfect image.

An image comes with metadata such as publish date, categories, tags, and who contributed it
For the purpose of this hackathon, we took and edited photos in-house to illustrate the concept. In the product’s future state, we have planned for a mechanism to source licensed images from local photographers.
“Search-by-content”: We recommend you the perfect photo for your content
When we were doing research, we found out that some officers might have difficulty deciding what kind of image to look for in the first place. We built a feature that helps to translate your content to photos. Users can simply add in a title and GovStock suggests photos for the content.

Enter the title of your content and govstock will recommend you images
Solving for quality via supply
Ultimately, it’s the quality of images we’re hoping to improve. Through our research, we recognised that branding standards and communications strategies vary across agencies, so the easiest place to solve for government-wide quality is supply. That’s why our system is designed to allow cross-agency licensing of contracted photos and usage tracking, as well as provide advice to photographers on the type of imagery in demand across government and solicit license-free contributions from members of the public. All of these features are intended to boost supply of usable images.
Research & approach
We assumed a number of things going into the hackathon:
Images on government collateral (and from similar sectors such as finance) impact citizen trust and engagement in communications
Government comms teams use stock images regularly
Comms teams have trouble finding quality / relevant stock imagery
Comms teams would be willing to share images across agencies if they could
We validated our assumptions and gained in-depth understanding of the challenges faced by government comms teams (detailed above) through interviews with comms officers from four government agencies, as well as marketing and branding experts from the private sector. These included:
Agencies: CPF, MOH, MSE, MTI
Ibrahim Ramthan - Former Assistant Director, Futures, Strategy and Plans Division (Intelligence and Futures), MTI
Chee Kang Xiong - Senior Assistant Director (Digital and Visual Communications), Communications and 3P Partnership Division, MSE
Dannon Har - Senior Manager (Branding, Digital and Visual Comms), Communications Division, MOH
Astin Lee, Senior Deputy Director, Media & Publicity Dept., CPF
Private sector experts
Anton MacDonald - Global Brand Manager, Mizuho Financial Group
Ryan Ducie - Former Group Head of Marketing, MoneyHero (Hyphen) Group
Mark Johnson - Former Senior Adviser @ Amazon Pay (Communications, Public Policy, Public Affairs)
Grant Petrie - Video & Animation Team Lead, RR Donnelley

Lots, lots, and lots of virtual sticky notes
Feedback
On demo day, we were able to gather feedback on the first iteration of our platform. Interviewees from agency comms teams shared the following:
“If it’s AI proposing the images, and you can search the whole of government’s photos in this library—that’s impressive.” - MSE social media officer
“Provenance is clearly labeled—it can help get people on your side for your pitches. Type of images are very accurate to what’s being used.” - MTI comms officer
Decision-making process
We then explored various approaches to the challenges they highlighted, including:
Stock library
AI image generator
Tools that could advise or auto-adjust images to be more consistent with other content or branding guidelines
AI that could make image & styling recommendations based on how popular sites with similar content had visualized theirs
Visual training tool for comms officers
Internal approval system for images
We discarded AI image generation as a main focus based on feedback in our user interviews that highlighted concern over AI undermining trust in the veracity of public communications.
Early on we were able to develop a wireframe stock library with AI-driven content-based recommendations and trial it with 3 agencies, to positive reviews. Based on this feedback and an in-depth discussion & vote on the various ideas surfaced, we fixed on the current solution.
We noted in our interviews, however, that some agencies have stricter parameters around the images they use, or source images before written content is concrete enough to use content-based search. We therefore added a keyword-based search.
Comms officers also told us that agencies would often contract photographers to take stock imagery—or create their own—but there were barriers to sharing such images across government. Restrictive licensing contracts mean only the contracting agency can use the imagery, and the risk of using photos or models in conflicting content is also a concern. Based on this input, we determined flexible licensing clauses and provenance / usage tracking would be key features of the system.
Challenges
Usage tracking has emerged as a potential challenge. We are exploring numerous solutions, such as voluntary reporting, download tracking, digital watermarking, reverse image search tools + web scraping, a built-in approval process, and others.

Taking our own photos in the OGP office to test out the model
Next steps
The prototype illustrates the concept of a curated stock photo library of Singaporean models and scenes that are fit for government communications, with both keyword- and content-based searches.
For GovStock to move beyond the concept phase, there are a few next steps:
Acquiring stock photos from local photographers
We aim to find agency partners that can help set up flexible licensing and/or incentive structures with local photographers to contribute to GovStock’s library and get fairly compensated. We will also draft appropriate terms & conditions to enable non-contracted photographers to contribute their work. A section of the platform dedicated to photographers will track and publish the top search terms on the site to help prompt them to take and contribute images with high value.
Tracking usage & models
To de-risk cross-agency usage of images, we will explore methods to track photo and model content associations.
Auto-tagging and vetting
We have plans to introduce AI-based auto-tagging for uploaded photos based on facial recognition (already introduced by PinPoint) and image recognition in order to streamline the contribution process. We also recognise accepting contributions from non-contracted photographers means we will need to set up a vetting system.
Photo licensing / purchasing system
Not all of our images will be free. To ensure proper use of and payment for photos, we plan to introduce a license tracking system and a photo purchasing function. In addition, some agencies may been keen to restrict the usage of some of their content due to sensitivity around depictions of people or places, yet still upload this content to make use of our search and tracking capabilities. This would necessitate a more nuanced licensing system.
Testing out demand with agency users
Using the concept prototype, we aim to understand whether the solution meets communications officers’ needs.
Expanded possibilities
Ultimately, images are one component of government communications. GovStock could form a part of a more comprehensive tool with other possible features, such as:
Integration with Isomer
Government-wide or agency-specific branding guidance
Text content advisory
Translation tools, such as Isofluent
Comms approval system that trains AI advisor, ultimately reducing approval time
So what?
If you’re a government comms officer: Come talk to us! Find us at Hack for Public Good Demo Day or complete our interest form. Tell us about your visual communication challenges and explore how GovStock can help solve them!
If you’re a photographer: We want to speak to you too! Share any insights you have about shooting and monetising stock photography or working with government agencies. And put your photos on GovStock!
If you’re a member of government or the public: Tell us about your experience with government visual communication. What works well? What can be improved?
The team behind the scenes

Humans of GovStock | From left to right: Xian, Sam, Se Hyun, Azhar, Sean, Prakriti
📷 GovStock is brought to you by:
Sean M, project lead and photographer in residence
Azhar S, production lead and photographer
Prakriti B, engineer
Samantha L, designer and engineer
Xian L, designer and editor
Se Hyun P, designer
❤️ Special thanks to:
Darren N, researcher
Models from Open Government Products - Sam, Nat, Aiko, Sufyan, Kenneth, Huaying, Jan, Jinhui, Alwyn, and Hena