Best AI Prompts for User Persona Visualization with This Person Does Not Exist

TL;DR

This Person Does Not Exist generates photorealistic human faces that can serve as the foundation for visual user personas, making abstract personas more emotionally tangible for design teams
The tool works best as a face-generation engine — you generate a face, then pair it with persona context to build a complete visual persona
Variety generation is the primary workflow — generate multiple faces per persona concept and select the most resonant one
This Person Does Not Exist faces are ethically cleaner than stock photos because they represent no real person, avoiding representation and consent issues
The best persona visualizations combine AI-generated faces with contextual design — the face alone does not create empathy, the face in context does

Introduction

User personas are one of the most widely used research-to-design translation tools in the product development toolkit. The concept is simple: synthesize research about your users into archetypal characters that your team can reference when making design decisions. The problem is that most persona documents fail at the one thing that makes them valuable: creating genuine empathy.

A text-based persona — name, age, job title, goals, pain points — engages the rational part of the brain. It is useful information, but it does not trigger the emotional response that changes how someone makes a design decision. When a product manager looks at a feature proposal and thinks “our users would want this,” a text persona does not give them a face to put to that decision. A visual persona does.

This Person Does Not Exist (TPDNE) is a specialized AI face generator created by Chip Hall. Unlike general image generators like Midjourney or DALL-E, TPDNE is specifically trained to generate photorealistic human faces that are indistinguishable from real photographs. The interface is intentionally simple: refresh the page, get a new face. No prompts, no parameters, just pure face generation.

This simplicity is both a strength and a limitation. TPDNE excels at generating high-quality, bias-tested faces quickly. It cannot control demographics, expression, or style — you get what the model generates. This guide covers workflows that work within those constraints to build genuinely useful visual personas.

Understanding This Person Does Not Exist
The Visual Persona Workflow
Generating and Selecting Faces
Pairing Faces with Persona Context
Building Complete Visual Persona Cards
Using Visual Personas in UX Research
Ethical Considerations
Limitations and Workarounds
FAQ

Understanding This Person Does Not Exist {#understanding-tpdne}

This Person Does Not Exist uses StyleGAN, a generative adversarial network architecture developed by NVIDIA, trained specifically on high-quality portrait photographs. The result is a model that produces faces that look remarkably real — the skin texture, the hair detail, the subtle asymmetries — because it learned from actual photographs.

The key characteristics of TPDNE:

No prompt interface. Unlike Midjourney or DALL-E, you cannot type a description. You refresh the page and get a random face. This is a limitation if you need a specific demographic, but it is a strength in one important way: the model defaults to a wide variety of faces across age, ethnicity, gender presentation, and style. There is no prompt injection bias because there is no prompt.

Refresh-based generation. Each page refresh generates a new face. For persona work, you typically generate 10-20 faces per persona type and select the most resonant one.

High resolution and quality. The generated faces are high enough resolution to use in presentations, print documents, and design artifacts. The model is well-maintained and produces consistent output.

Single face per image. TPDNE generates one face at a time, centered and cropped. This is perfect for persona work, which typically needs one primary face per persona card.

The Visual Persona Workflow {#the-visual-persona-workflow}

Building visual personas with TPDNE involves four steps that leverage the tool’s strengths.

Step 1 — Define your persona types. Before generating faces, you need to know what persona archetypes you are building. Use your existing research to define 3-7 distinct user types. For each, document: age range, profession type, lifestyle context, emotional quality (stressed, confident, curious, etc.), and tech relationship (early adopter, pragmatic, skeptical).

Step 2 — Generate candidate faces. For each persona, generate 15-20 faces by refreshing TPDNE. As you generate, keep notes on which faces resonate with the persona description you wrote in Step 1. Most will not fit — this is normal. The goal is to find the one or two that feel right.

Step 3 — Select the strongest face. From your candidates, select the face that best represents the persona. Consider: does this face look like someone who would use my product in the way my persona describes? Does it trigger the right emotional response? Does it feel like a real person?

Step 4 — Build the complete persona card. Pair the selected face with persona context: name, role, goals, frustrations, a brief narrative description, and a representative quote. The face is the anchor; the text provides the substance.

Generating and Selecting Faces {#generating-and-selecting-faces}

Because TPDNE has no filtering controls, face selection is an iterative process. Here is how to be systematic about it.

For each persona type, generate candidates and sort them into three buckets:

Strong fits (2-3 faces): These faces immediately match the persona in your mind. They look like someone who would say the things your persona says. Keep these.

Potential fits (5-8 faces): These faces are not immediately obvious but could work with the right context. Keep them — sometimes a face that did not initially resonate becomes perfect once you pair it with the right persona narrative.

Poor fits (discard): These faces do not match the persona type. Do not spend time on them.

Selection criteria for the final face:

Emotional resonance: does looking at this face trigger an empathetic response?
Representativeness: does this face look like someone who represents a significant user segment?
Distinctiveness: is this face clearly different from your other persona faces? (If all your personas look similar, the set will not be useful)
Credibility: does this look like a real person, not an AI artifact?

Pairing Faces with Persona Context {#pairing-faces-with-persona-context}

A face alone is not a persona. The value comes from the combination of image and substance. Here is how to build that combination.

Prompt (used with Claude or ChatGPT to build persona context):

I have selected a face for a visual user persona representing: [PERSONA TYPE — e.g., "a pragmatic mid-level marketing manager at a mid-sized company who is skeptical of new technology tools"].

Persona characteristics:
- Age range: [RANGE]
- Professional context: [JOB LEVEL, INDUSTRY, ORGANIZATION TYPE]
- Relationship with technology: [EARLY ADOPTER / PRAGMATIC / SKEPTICAL / RESISTANT]
- Primary work frustration: [THEIR MAIN PAIN POINT]
- What they need from tools: [THEIR CORE JOB-TO-BE-DONE]
- Decision-making style: [INDEPENDENT / CONSENSUS / HIERARCHICAL]

Generate a complete persona card for this person:
1. A first name — should feel representative of their background and era
2. A brief persona narrative — 2-3 sentences that paint a picture of a day in their professional life
3. A representative quote — something they might actually say that reveals their attitude toward their work and tools
4. Their primary goal when using [YOUR PRODUCT CATEGORY]
5. Their biggest frustration with tools like [YOUR PRODUCT CATEGORY]
6. How they would describe a good tool to a colleague

[FACE DESCRIPTION OR REFERENCE]

The persona card should make the abstract user type concrete. A face paired with genuine narrative context is what makes design discussions productive — the face triggers the emotional recognition, the text provides the substance for decisions.

Building Complete Visual Persona Cards {#building-complete-visual-persona-cards}

A complete visual persona card for UX use includes:

The face: The TPDNE-generated image, cropped cleanly and sized appropriately for your outputs (presentation decks, research documents, print).

Persona name: A representative first name that humanizes the archetype.

Demographic snapshot: Age range, role, industry, location — not exhaustive, just anchoring context.

Core narrative: 2-3 sentences that paint the picture of this persona’s working life and relationship to the problem your product solves.

Jobs to be done: The 2-3 primary tasks this persona uses your product to accomplish.

Pain points: The top frustrations this persona experiences with the current solution or workflow.

Tech relationship: How this persona approaches new tools — do they research extensively, try quickly, ask colleagues, trust brands?

Representative quote: A first-person statement that captures this persona’s attitude.

For slide/presentation use:

Format this persona as a one-slide persona card suitable for executive presentations. Include:
- The face image at top (half the slide)
- Name and role as a header
- 3 bullet points max: primary goal, key frustration, tech approach
- One representative quote
- No more than 5 seconds of reading time for the text

Using Visual Personas in UX Research {#using-visual-personas-in-ux-research}

Visual personas add value at multiple stages of the UX research and design process.

In stakeholder alignment: Show persona cards in discovery sessions to anchor discussions about users. Stakeholders often have different assumptions about who the user is — a shared visual reference creates a common starting point.

In design critique: Start design reviews by having each participant identify which persona the design decision benefits. “Who is this screen designed for — Priya who is time-poor and skeptical of complex tools, or Marcus who is an early adopter who wants deep customization?” These discussions produce better design decisions than abstract “users might prefer.”

In user story mapping: Attach persona faces to user story cards. When prioritizing features, teams can viscerally feel who benefits from each feature and make prioritization decisions from that empathetic reference.

In research synthesis: When synthesizing user interviews, assign each participant to a persona type and generate a visual persona for each type. This helps the team remember research participants as people, not data points, which reduces confirmation bias in interpreting findings.

Ethical Considerations {#ethical-considerations}

TPDNE-generated faces raise fewer ethical concerns than stock photography for persona work, but some considerations remain.

No real person is represented. Unlike stock photos, which depict real humans who gave consent to have their image used, TPDNE faces represent no actual person. This eliminates concerns about consent, representation, and the commodification of real people’s images.

Bias in the training data. Like all generative AI, TPDNE’s training data reflects biases in the source photographs. The tool has no demographic controls, which means generated faces may skew toward certain demographics if the underlying training data skewed that way. For serious persona work, be deliberate about generating faces across the full range of demographics your actual users represent.

Disclosures in client work. If you are using AI-generated personas in client-facing deliverables, disclose that the faces are AI-generated. Most clients do not have an issue with this, but presenting AI faces as research-based representations of real user segments without disclosure is misleading.

Limitations and Workarounds {#limitations-and-workarounds}

TPDNE’s simplicity is its biggest limitation. You cannot control:

Demographics: You cannot generate specifically an Asian woman in her 50s or a Black man in his 20s. You get what the model generates.
Expression: You cannot control whether the face looks happy, stressed, or focused. The model generates mostly neutral-to-positive expressions.
Style: You cannot generate an illustrated or flat-design face. All faces are photorealistic.

Workaround for demographic specificity: Generate enough faces that statistical probability gives you the demographics you need. For 20 generated faces, you will typically get a reasonable distribution across age, gender, and ethnicity. Select from that distribution for the personas you need to represent.

Workaround for expression: Select faces that have an inherent emotional quality in their neutral expression. Some faces look more stressed, more confident, or more curious even in a neutral expression. This is somewhat random but can be leveraged with enough candidates.

For illustrated or flat-style faces: TPDNE is not the right tool. Use Midjourney or a illustration-focused AI tool if you need non-photorealistic persona visuals.

FAQ {#faq}

How is TPDNE different from Midjourney for persona visualization?

TPDNE is specialized for photorealistic faces only — it does one thing and does it very well. Midjourney is a general image generator that can create faces in any style, in any context, with any parameters you specify. The trade-off is control versus quality. Midjourney gives you control but requires skilled prompting to get consistent, high-quality results. TPDNE gives you high-quality faces instantly but with no control over demographics or expression. For most persona work, TPDNE is faster and produces better faces, but Midjourney is better when you need specific demographic representation.

Can I use TPDNE faces in commercial design work?

Yes. The faces generated by This Person Does Not Exist are free to use without attribution. The images are synthetic and do not represent real people, which eliminates most commercial licensing concerns that apply to stock photography. However, if you are using them in client work, good practice is to disclose that they are AI-generated rather than photographs of real users.

How many personas should I generate for a product?

The research consensus is 3-7 personas per product. Fewer than 3 often means you are overgeneralizing, and more than 7 makes it difficult for design teams to hold them all in mind during decision-making. Focus on the 3-5 primary user types that represent the majority of your user base, plus 1-2 edge case personas that drive critical design requirements for specific accessibility or use case needs.

What if I need diverse representation across specific demographics?

If you need specific demographic control over your persona faces, TPDNE’s random generation may not be sufficient. In this case, use a tool like Midjourney with explicit demographic prompting, or use multiple generations from TPDNE and be systematic about selecting from your generated pool to ensure demographic coverage across your persona set.

How do I avoid all my personas looking like they are from the same demographic?

Generate faces in batches and sort by demographic characteristics before selecting. TPDNE’s training data does have some demographic skew, so actively curating across your generated faces is important to build a representative persona set. Document the demographic rationale for each face selection so the choice is intentional rather than default.

Conclusion

This Person Does Not Exist is a specialized tool that excels at one thing: generating photorealistic faces quickly and without bias introduced by prompting. For visual persona work, this simplicity is a feature. The workflow is fast, the faces are high quality, and the ethical surface area is low.

Key takeaways:

Generate 15-20 faces per persona type and sort into strong fits, potential fits, and discards
Pair every face with genuine persona context — the face alone does not create empathy
Use visual personas as empathy artifacts in design critique and stakeholder alignment
Be deliberate about demographic diversity across your persona set
Disclose AI generation in client work

Your next step: define your 3-5 primary user personas from your existing research, then spend 10 minutes generating faces for each using TPDNE. Select the strongest face per persona and pair it with a narrative card. Compare how your team responds to the visual version versus the text-only version — the difference in engagement is where the value lives.

Best AI Prompts for User Persona Visualization with This Person Does Not Exist