The Challenge:

Like most digital products, Folktale often needs rich visual assets—images, videos, and more—to illustrate how our product is used and in what contexts. We use these visuals on our website, social media, and in email communications.

Since our platform is fully private and stories are owned by the contributors, we never use their stories outside the agreed consent. So, when we need promotional images or video content, we partner with local photographers and videographers in the communities we serve. These local creatives are essential; they understand the culture and can produce visuals that are representative and respectful.

We provide a creative brief that outlines our goals, the intended message, and the mediums for display. This brief typically includes specifics such as the age, ethnicity, and activities of the people we aim to represent, along with the setting—whether it’s a bustling city or a quiet village.

However, this approach presents challenges. It requires significant time, coordination, and resources. Language barriers and creative interpretations can lead to visuals that don’t fully meet our needs, and managing inconsistent outputs across multiple locations can be complex.

What I Hoped to Achieve:

First, I want to emphasize that I have no intention of taking work away from local creatives. While AI-generated imagery might offer some cost savings, empowering creative communities is a key priority for me, deeply aligned with both Folktale’s and my own values.

However, many of the communities we work with are remote, underserved, and sometimes marginalized, making it difficult—or even unethical—to obtain and share real-world images.

Could AI-generated imagery help us represent the kinds of environments and people engaging with our product, while preserving confidentiality?

Could it assist us in crafting clearer briefs that more effectively convey our expectations, reducing the chances of misinterpretation and minimizing the need for reshoots?

The Experiment:

I began experimenting with Leonardo.ai to generate photorealistic images, about six months after an initial attempt with DALL-E that didn’t yield the results I was hoping for. The improvements in such a short time were remarkable. I was curious to see if AI could streamline our workflow while preserving the cultural accuracy and authenticity that are essential to us. Instead of trying to convey to a photographer the depth and nuance I envisioned in a person’s face, could I just show them an example?

I started by creating a detailed prompt—similar to the creative briefs we provide to local photographers. This prompt included cultural specifics, like:

• A middle-aged Asian woman in a rural Cambodian village

• Documenting community activities for a water and sanitation project

• Specific emotions we wanted to capture, such as a sense of empowerment or confidence

In seconds, I had a batch of photorealistic images. My first reaction was, "Wow." 

But on closer inspection, the results were mixed.

On the positive side:

• AI generated realistic images almost instantly, saving us time in sketching and searching through stock libraries.

However, there were challenges:

• AI sometimes misinterpreted prompts, inaccurately capturing cultural nuances (for example, confusing Papuan and Indonesian features).

• While some images looked impressive, others had technical flaws—like people with three arms or distorted hands.

• The AI struggled with human emotion and subtle concepts like “confidence” or “empowerment”; many faces had a strangely "hollow" expression.

This experiment left me with more questions than answers:

How do we ensure that AI-generated images don’t perpetuate harmful stereotypes?

What are the ethical implications around consent if AI-generated faces closely resemble real people?

As an organization committed to amplifying community voices, these are questions we must consider carefully, and they guide our cautious approach to AI integration.

What I Learned

Generative AI provides a quick, efficient way to create visuals, streamlining tasks like generating reference images or curating mood boards to guide local creators. However, as we’ve seen, generative AI currently lacks the cultural nuance and emotional sensitivity that authentic storytelling requires. Will it ever reach that level? Who knows, but the progress so far is astounding.

This highlights an important point for us at Folktale: while AI offers exciting possibilities, it can’t replace the relationships we’ve built with community creators. Instead, we see AI as a complementary tool—one that facilitates our workflow and streamlines communication with collaborators through various mediums, such as visuals. I envision AI supporting early project stages by accelerating template creation, simplifying repetitive tasks, and providing recommendations for tagging and labeling. However, I can’t see it replicating the lived experience and cultural insight that human collaborators bring.

But—who’s to say it won’t one day? If AI is driven and refined by human input, when might we reach a level of parity?

Considerations and Tips for Using Generative AI Tools

After experimenting with AI tools over the past year, I’ll continue exploring their potential as they evolve. At Folktale, we’re also developing a policy to guide how we use AI more broadly, which we’ll share when complete.

For now, if you’re considering AI tools in your own work, here are some guiding principles:

1. Safeguard Personal Information – Never use sensitive or personal data to train AI models.

2. Consider the Implications – Our work has a real-world impact, so approach it with care.

3. Think of AI as a Tool – Rather than a replacement, consider AI as a way to facilitate your workflows.

4. Prompts Matter – Be specific and precise in your prompts; crafting effective prompts takes thoughtful consideration to achieve the best results.

Next Steps

In early 2025, we’ll commission a new set of materials, collaborating with creatives globally. With our new knowledge, the brief will include Gen-AI-generated images as guides for framing, composition, and overall style. I’ll also write a follow-up blog comparing these outputs and sharing feedback from the creatives who will receive them.