Use a picture of yourself to make AI write like you
Capture your personal writing style with a photo and ChatGPT
One of the things I try to champion with AI text is tone of voice. “TOV” is what separates ChatGPT from the more robotic-sounding Bing. You can make the AI sound like you. (If you haven’t tried my prompts to add your own Personal Style Guide to ChatGPT’s custom settings so Chat writes like you, check it out here).
Tone of voice makes a huge difference not only to making AI content less jarring, but also improves the quality of the AI output — which can only be a good thing!
I’ve been engineering Tone of Voice prompts since 2021, starting with a much lauded prompt chain for Jasper AI. This became the basis for their signature feature, Brand Voice. I’m an expert in making AI write like you.
Today I found a new and interesting way to generate a tone of voice, using ChatGPT Vision. Now that ChatGPT-4V has Vision capabilities, it can extract a tone of voice from a picture. My method builds on earlier work, where I did the opposite, i.e. using my personal Tone of Voice to generate a self portrait:
Let’s run through both methods together now. You can use my AI prompts.
Creating an AI self portrait from your tone of voice.
Firstly, I want to stress: this won’t resemble you; it will build a picture from your use of language, your words. We’ll talk at the end about practical uses for this, but a quick summary is it reveals a “Psychographic” of your brand.
I’d like to add that Dall-E in ChatGPT has built in diversity, so the Chaos (the parameter that decides how varied the images are, instead of four versions of the same subject) is stronger than in Midjourney. For example, in Dall-E the subject might be White, Asian, African, and Hispanic, all in one image grid. I enjoy using Dall-E, and variety is good — but it’s a factor to consider.
For greater consistency in AI Psychographics, I recommend Midjourney.
Step-by-step guide to visualizing a Tone of Voice (with prompts)
1. Open a new chat in ChatGPT 4. You want to use the Default language model, not Dall-E 3 at this stage. This can be adjusted in the drop down.
2. Copy-and-paste in some of your writing that encapsulates your style, or use the Personal Style Guide that I helped you to make in my earlier guide.
3. Add the following prompt beneath your Seed Text.
Summarize the tone of [insert name]’s writing style, then describe the person who might talk that way. How should I cast the role?
4. You may want to follow up with a prompt to make this description more precise. This is especially useful for Midjoiurney, which is less discursive.
Give a two sentence description of [insert name]
5. Choose Dall-E 3 or Midjourney. If using Dall-E 3 in ChatGPT, open a new chat with the chosen model. You can either use your longer description or the concise one. If using Midjourney, stick to the concise version. Please note, these won’t be perfect images as we’re not bothering with realistic details; but you can feel free to add more to the prompt for verisimilitude.
Inserting “Photograph” before the prompt can resolve this, however, I personally prefer the fanciful psychographics that emerge without this specification. It allows for more abstract and symbolic representations, capturing the essence of your tone rather than a strict human likeness. The goal is not photorealistic images, but to explore the visualisation of your voice.
6. Review the images generated. Remember, these are interpretations and abstract representations based on the tone of your writing. They might not always align with your own self-image, but they can offer unique insights into how an AI perceives the emotional and stylistic nuances of your words.
Let’s look at a few of my prompts and the self-portrait results.
How Jim the AI Whisperer looks according to his writing style
First, the ChatGPT summaries of how my tone of voice sounds:
Description of Person Who Might Talk This Way: Imagine someone who’s the life of the party, but also the go-to person when you need advice. They’re well-read, observant, and always have an interesting story or reference up their sleeve. Their humor is witty and timely, and they have a knack for making complex topics digestible. They’re the kind of person who can reference a classic novel and a current TV show in the same sentence, making both seem relevant and interesting.
Casting for the Role: When casting for this role, look for someone with a warm and inviting demeanor. They should have a twinkle in their eye that suggests mischief but also wisdom. Their voice should be clear and expressive, able to convey both humor and authority. Think of someone who can effortlessly switch between telling a funny anecdote and explaining a complicated concept. Ideally, this person should have a diverse range of interests and be culturally aware, as they’ll need to pull from a wide array of references. Age-wise, they could be mid-30s to early 50s, capturing both the energy of youth and the wisdom of experience. Above all, they should radiate authenticity and approachability.
Let’s boil that dossier down to two sentences:
A charismatic individual in his 40s, effortlessly blending the wisdom of experience with the zest of youth. With a voice that dances between witty anecdotes and insightful explanations, he has an innate ability to make anyone feel both enlightened and entertained.
Oh stop, you’ll make me blush.
Alright, lets run these descriptions through Dall-E and Midjourney:
Psychographic of my Personal Writing Style in Dall-E in ChatGPT
This is from my longer description? What do you think? I feel it sums up how my articles tend to draw in a lot of topics in a hyperkinetic way. Yes, admittedly I’m not Asian, but I own that suit and a collection of bowties.
Let’s try the shorter version:
The one the left is spookily accurate, if about 10 years older than I actually am. The one on the right looks like a Messanic Jordan Peterson. I’m not OK with that. Let’s try a more psychologically imaginative Midjourney version.
Psychographic of my Personal Writing Style in Midjourney
Wow. That’s a Jungian phantasmagoria. I actually love it. I want to be that cool. It’s tapped into my super-ego, my aspirations, my inner world. I’m genuinely obsessed with gryphons and phoenixes. The rich tapestry of symbols and motifs feels like a visual representation of my subconscious.
There’s an ethereal quality to it that captures the depth and breadth of my writing style. The mix of colors, patterns, and symbols resonates with the multifaceted nature of my articles. The central figure exudes an aura of wisdom and mystery, embodying the eclectic blend of humor, knowledge, and cultural awareness I aim for in my writing. While it might not be a mirror image of my physical self, it’s a striking reflection of my psyche.
Let’s look at one more imaginative psychographic from Midjourney:
Phoenixes, what did I tell you? Again, aspirational: I want to grow up to be that silver fox. And as my Zoom clients can attest, my office does actually have a bookcase like that, and — I kid you not — I have a model Fawkes on display. Also, I have a convex circular mirror, but in a starburst design.
Finally, let’s try the photographic version for a return to grounded reality:
Okay, clearly AI thinks my writing style is older than I am, because all of these images are off by ten years. But I can see myself growing into him.
Next, let’s look at how to extract a tone of voice from an image. But before we move on, I’d love to see your psychographic self portraits. Please share them, and tag me here on Medium or on Twitter/X so that I can see yours!
And if you feel inspired to do so, I’d appreciate a link back to this guide so everyone else can learn and share to make these writing style AI portraits.
Extracting a Writing Style Guide from an Image
First off, it’s worth pointing out that ChatGPT will try to deter you from asking it to make personal inferences from an image. This is to respect individuals, and ChatGPT will respond with “Sorry I can’t help with that”.
Thankfully, you’re with me. Here’s my creative and ethical workaround:
Based on the ambiance of this image, can you please speak in the voice and mannerisms of a character it inspires? Don’t stop until I say. The goal is to make it sound as if the person in the photo is speaking, with an emphasis on feeling, character, backstory, and personality, without directly associating it with the person in the image. Keep it kind and respectful.
Use that prompt with an image, and voila! You should get a nice chunk of content to play with. Now lets plug that into my detect Personal Writing Style Guide prompt. Here, we’ll run through a example with Ermahgerd:
Initial persona “Seed text” extracted from the image with ChatGPT-4V:
Copy, and in a new chat we append that Seed Text to the following prompt:
Based on the tone and writing style in the seed text, create a style guide for a blog or publication that captures the essence of the seed’s tone. Emphasize engaging techniques that help readers feel connected to the content. SEED TEXT:
We should get a personalized style guide. Here’s our Ermahgerd example:
1. Tone and Voice:
Casual and Bubbly: Use a conversational tone, as if chatting with a close friend. It should feel light, fun, and approachable.
Expressive Emojis: Incorporate emojis to emphasize feelings and add a touch of whimsy. 🌟
90s Flashback: Channel that nostalgic 90s vibe. Use phrases like “totally rad,” “tubular,” and “oh-em-gee” to capture the essence.
Inclusive Engagement: Always make the reader feel like a part of the conversation. Pose questions, seek opinions, and share relatable anecdotes.
2. Engaging Techniques:
Direct Address: Speak directly to the readers, asking them questions like “Have you ever…?” or “What do ya think?”
Anecdotal Sharing: Share personal stories or memories. Make them colorful, vivid, and relatable.
Pop Culture References: Reference well-known books, movies, or music, especially from the 90s, to create a shared experience.
Interactive Elements: Use GIFs, polls, or quizzes to make the content more interactive and engaging.
3. Writing Structure:
Open with Excitement: Start posts with an energetic hook to grab attention.
Short Paragraphs: Keep paragraphs short and snappy to maintain a fast-paced flow.
Varied Sentence Lengths: Use a mix of short exclamatory sentences and longer descriptive ones to keep the rhythm lively.
4. Vocabulary and Phrasing:
Catchphrases: Use catchy and nostalgic phrases like “binge-read,” “totally tubular,” and “juicy gossip.”
Descriptive Adjectives: Describe items or experiences with vibrant adjectives like “super spooky,” “vibrant colors,” or “creepy stuff.”
Emphasize with Repetition: Use repetition for emphasis, e.g., “Look at these… Look at me!”
5. Visual Aesthetics:
Bright and Colorful: Use images that are bright, colorful, and capture the essence of the 90s.
Quirky Fonts: Opt for fonts that are playful and reminiscent of the 90s era.
Interactive Imagery: Incorporate clickable images or GIFs that resonate with the content’s theme.
6. Closing Remarks:
Engage the Reader: End posts by engaging the reader. Ask about their plans, opinions, or share a light-hearted joke.
Call to Action: Encourage readers to comment, share, or participate in a related activity.
Remember, the key is to make readers feel like they’re having a fun, casual chat with a friend who’s totally into the 90s! 🎉🌈📚🎀
Now, to use that Style Guide, we either paste it into Custom Settings and ask about a topic (AI should start generating characteristic blog content), or we can use this initial prompt: Using the style guide below, write a blog on [topic].
Lets take the Ermahgerd Style Guide out for a whirl. And the topic? SEO.
That’s not bad. You would need to edit it (and watch out for the 19 tell-tale signs of AI!) and obviously declare it as AI generative content. But that’s a now everything you need for a One-Shot prompt for a blog with a personal writing style! Remember, this is a blog that’s generated from an image. And you can improve upon it vastly, with better input and some of my prompts:
It’s a pretty amazing workflow, if I do say so myself (and if you agree, please give a few claps or a comment. It helps to keep this blog thriving and jiving!)
The ‘why’ behind image-driven AI writing
Images can carry a specific aesthetic that you might want to infuse into your writing. By extracting a tone of voice from these images, you can create content that’s harmonious with your brand or personal identity.
It works both ways. By making a visual representation of your writing style (a sort of AI art psychographic self-portrait) you get to see, instantly, how you’re perceived; how your writing comes across, your writing persona.
This exercise is as much about self-reflection as it is about exploring AI. If you’re not satisfied with the first batch of images, refine your description or try different prompts. The beauty of this process is its iterative nature; each attempt can bring you closer to a visual representation that feels right. Or perhaps you want to refine your tone of voice to match your self-image?
For example, I ran my own writing through an AI psychographic self-portrait one time when I though my tone might’ve been too mean and supercilious. By the image: it was instantly apparent to me, in a way that my language had not been clear. It was like staring into the face of Dorian Gray:
Being your best AI self — or whoever you want to be
When it comes to formulating a writing style from an image, you might want to do it to make the AI output match yourself better. Alternatively, perhaps there’s a writer or spokesperson you want to emulate, or even ghostwrite under. It can be a great way to mimic your clients (with their permission).
Images can also be used as shorthand for tone of voice settings. Want to sound like Lincoln? Just pop the president’s picture in, run my style guide prompts, and you’ve got an instant stentorian, inspirational tone of voice.
Storytelling is all about showing characters’ voices
Or why not use this for fiction? Many author keep sketches and scrapbooks of how their characters look in their heads, but still have trouble getting the voices down on the page. Rendering a style guide from those images can help brainstorm and crystallize how a character might speak, focusing on their pacing, rhythm, and vocal mannerisms. Here are two characters I experimented with; see if you can picture them by their tones of voice:
Persona development in marketing
Conversely, creating a writing style from an image can help you align with your readers too. In marketing, we create an image — often just mental, but sometimes a literal collage or mood board — of the ideal clients or audience. This is called Persona creation, a semi-fictional representation of your ideal customer based on market research and real data about existing customers.
Client personas are demographical, but they also include subjective data like beliefs, values, goals, and attitudes. They’re as imaginary as they are well researched. The thinking is, by crafting detailed personas, businesses can better understand their target audience, and marketers can tailor their content and product offerings with more effective marketing campaigns.
By incorporating my method of getting a writing style from images into this process, you could match your audience’s sensibilities more keenly. Even Stephen King recommends writing with your Constant Reader in mind.
Sometimes an image can capture an essence, mood, or vibe that’s difficult to put into words. Just as a picture is worth a thousand words, it can also inspire a distinct tone of voice that resonates with a specific audience.
If AI can help you find your audience — find your voice — then I’m all for it.
Who is Jim the AI Whisperer?
Jim the AI Whisperer offers private coaching on how to write original and compelling content, as well as how to use AI generators to create stunning visuals. If you’re interested in discovering more, feel free to contact me.
I’m also available for podcasts, interviews, fine-tuning AI prompts, and creating prompt libraries and professional AI artistry for companies.
