Nano Banana LinkedIn Headshot Prompts: The Manual Method and the Shortcut

Last updated: July 15, 2026.

If you've typed "turn this selfie into a professional headshot" into Gemini, you've already used Nano Banana. It's not a separate app, just the nickname that stuck for Google's image model after it showed up in the Gemini app's tools menu, and it's become the default way a lot of people are making their own LinkedIn photos in 2026.

This guide covers both sides of that: how to do it yourself in the Gemini app with prompts that actually work, and where the manual approach runs into friction that a purpose-built tool skips entirely. We're not going to pretend the manual method doesn't work, because it does. We're going to show you exactly how, and then be honest about where it gets tedious.

What Nano Banana actually is

Nano Banana is the nickname for Gemini 2.5 Flash Image, Google's multimodal image generation model, released in 2025. It reads an uploaded photo the same way it reads any other input, as data to reason about, and generates a new image conditioned on that photo plus your text prompt in a single step. No separate face-extraction pipeline, no per-user training.

As of June 2026, Google shipped a second, more capable model built on Gemini 3 Pro, officially called Gemini 3 Pro Image and nicknamed Nano Banana Pro. It adds legible text rendering inside generated images, native multi-turn editing, and up to 4K output, at a listed API price of $0.134 per image. If you're using the free Gemini app, you'll see both a "Fast" option (the original Nano Banana) and higher-tier options depending on your Google AI subscription. For a plain headshot, the original Nano Banana is more than capable, Pro's advantages (text rendering, ultra-high resolution) mostly matter for other use cases.

FreeHeadshot.org runs the original Gemini 2.5 Flash Image model in production. We're evaluating Nano Banana Pro for a future upgrade but haven't shipped it yet, and we'd rather tell you that plainly than let you assume otherwise.

How to do it yourself in the Gemini app

1. Open the Gemini app or Gemini on the web.
2. Select "Create images" (the banana icon) from the tools menu.
3. Upload your reference photo using the attachment icon, then add your prompt in the same message.
4. Pick a model tier if you have the option; "Fast" is the original Nano Banana and is sufficient for a headshot.
5. Send it. Generation typically takes a few seconds.
6. If the result isn't right, reply in the same thread with a follow-up instruction ("make the background a warmer gray," "less smile") rather than starting over. Nano Banana supports iterative editing on the same image within a conversation.

That's the whole mechanical process. The part that actually determines your result is the prompt, which is what the next section is for.

Why the wording of your prompt matters this much

Two prompts that mean the same thing to a human can produce very different results from an image model, and the gap usually comes down to vocabulary.

"Make me look professional" tells the model almost nothing concrete. It has to guess what "professional" means to you, and it'll default to whatever "professional headshot" looks like in its training data on average, which tends toward generic. "Navy blazer, white shirt, softly blurred office background, soft key light from the left, 85mm lens look" gives the model six concrete, photographic decisions instead of one vague one. Photography and lighting vocabulary works better than personality adjectives for a structural reason: terms like "soft key light" or "shallow depth of field" map to specific, learnable visual patterns in a model's training data, while "confident" or "handsome" don't correspond to anything visually consistent, so the model falls back on its most generic, averaged interpretation, which is exactly the plastic, over-smoothed look most people are trying to avoid.

The same logic applies in reverse to what you leave out. If you don't specify skin texture, the model defaults toward smooth. If you don't specify an aspect ratio, it infers one from your source photo, which may not match what you actually need. Being explicit isn't about writing a longer prompt for its own sake, it's about not leaving decisions to the model's average-case default when you have a specific one in mind.

17 prompts that actually work

Structure matters more than length. The prompts below follow a consistent pattern: task, subject, styling, background, lighting, and camera language, in that order. Camera and lighting vocabulary ("85mm," "soft key light") gets meaningfully better results than adjectives about the person ("handsome," "confident"), because the model has more to work with photographically and less room to default to an airbrushed, generic face.

Corporate / LinkedIn standard

1. "Turn this photo into a professional LinkedIn headshot. Navy blazer, white shirt, no tie. Softly blurred modern office background. Soft key light from the left, 85mm lens look, shallow depth of field. Keep the person's exact face and features unchanged, just the styling and setting."

2. "Professional corporate headshot from this photo. Charcoal suit, light blue shirt. Plain light gray studio background, seamless. Even, soft studio lighting from the front. Sharp focus on the eyes, natural skin texture, no retouching that removes pores."

3. "Business casual headshot. Crew-neck sweater over a collared shirt. Background: a softly out-of-focus bookshelf. Warm, natural window light from the side. Candid half-smile, not a posed grin."

Executive

4. "Executive-level headshot from this photo. Dark suit, no tie, top button open. Deep charcoal seamless background. Dramatic but flattering side lighting, slight Rembrandt lighting pattern. Confident, direct gaze at the camera. Sharp, editorial quality."

5. "Senior leadership headshot. Tailored navy suit. Background: blurred glass office interior, city lights barely visible. Low, warm key light. Serious but approachable expression."

Creative / startup

6. "Casual professional headshot for a startup team page. Plain t-shirt or henley, no blazer. Background: exposed brick or concrete, softly blurred. Natural daylight from a window, slightly cool color temperature. Relaxed, genuine smile."

7. "Creative-industry headshot. Interesting but not distracting clothing texture. Background: a colorful but blurred mural or gallery wall. Directional light with a visible catchlight in the eyes. Playful but professional expression."

Black and white / editorial

8. "Black and white editorial headshot from this photo. High contrast, strong single-direction light source, deep shadows on one side of the face. Studio background, pure black or pure white. Sharp focus, visible skin texture, no smoothing."

Outdoor / natural light

9. "Outdoor natural-light headshot. Soft, overcast daylight (no harsh shadows). Background: greenery, softly out of focus. Warm color grade. Genuine expression, slight head turn rather than square to camera."

Authentic / minimal retouching

10. "Realistic professional headshot with minimal retouching. Preserve visible skin texture, pores, and natural asymmetry. Avoid airbrushing or beauty-filter smoothing. Soft, flattering but honest lighting. Plain neutral background."

Academic / researcher

11. "Academic headshot from this photo. Smart-casual clothing, no tie. Background: a softly blurred university office or bookshelf. Even, natural-feeling light, nothing dramatic. Approachable, thoughtful expression."

Real estate / client-facing

12. "Approachable client-facing headshot. Business casual, warm color palette in the clothing. Background: softly blurred bright interior or exterior. Bright, even lighting, genuine warm smile. Trustworthy, friendly tone overall."

Remote team / company directory

13. "Consistent team-directory headshot style. Solid mid-gray background, no texture. Even front-facing studio lighting, no dramatic shadows. Neutral, professional but relaxed expression. Square crop, centered."

Common fixes as follow-up prompts

14. "Keep everything the same but make the smile smaller and more closed-mouth."

15. "Same image, but change the background to a solid warm gray, no texture."

16. "Reduce the skin smoothing, I want visible pore texture and natural skin, not airbrushed."

17. "Keep the framing and lighting but change the outfit to a plain white shirt with no jacket."

A few things that trip people up in the Gemini app

The model tier matters more than people realize. If your account defaults to a "Thinking" or reasoning-heavy tier, generation can take noticeably longer than the "Fast" tier for something as simple as a headshot, since the reasoning step is built for harder prompts than "make this a professional photo." Switching to "Fast" is usually the right call for this specific task.

Uploading a screenshot instead of the original photo file compresses the image twice and visibly hurts likeness quality. Always upload the original photo file, not a screenshot of it.

The conversation history affects later generations. Because Nano Banana supports iterative editing within a thread, a long conversation with several unrelated image requests in it can start bleeding style choices between them. If a headshot attempt is coming out strange for no clear reason, starting a fresh conversation with just the reference photo and prompt often fixes it.

What uploading to Gemini directly means for your photo

Worth being straightforward about, since it's a real difference between the manual route and a dedicated tool. When you upload a photo directly to the Gemini app, it's subject to Google's general consumer data and Gemini app privacy terms, the same ones that apply to everything else you do in Gemini. That's a reasonable, well-documented policy, just a general-purpose one, not one written specifically around a face photo's particular sensitivity.

A dedicated headshot tool can (and should) be explicit about exactly what happens to that specific photo: how long it's kept, whether it's used for anything beyond generating your result, and whether a face embedding is retained after your session ends. On freeheadshot.org specifically, photos are processed in memory, never used to train a model, and deleted on a fixed schedule rather than kept indefinitely; full detail is on our security page. If you're uploading a face photo anywhere, checking what the specific tool says about retention, not just trusting a general privacy policy, is a reasonable five minutes to spend first.

What consistently goes wrong with manual prompting

Aspect ratio drift. Left to its own judgment, Nano Banana infers a composition from the reference photo, which means one generation comes back square, the next comes back portrait, and matching a consistent set for multiple platforms (LinkedIn's circular crop, a resume's rectangular photo, a company directory) takes manual cropping afterward. Explicitly stating an aspect ratio in the prompt helps but isn't always honored precisely.

Identity drift across attempts. Regenerate the same prompt three times and you'll get three subtly different faces, not three lighting variations of the same face. For a one-off image this doesn't matter. For picking the best of several, it means you're not actually comparing variations, you're gambling on which attempt kept the closest likeness.

No batch consistency. If you want the same look in three outfits or three backgrounds for an A/B test on your profile, each one is a separate generation with its own drift, not a consistent set.

The smoothing default. Nano Banana, like most image models, defaults toward smoother skin unless you explicitly fight it in the prompt (see prompt 13 above). Forgetting that instruction is the single most common reason a manual Nano Banana headshot ends up looking a little too polished.

None of these are model failures exactly, they're the tradeoff of a general-purpose tool being asked to do a specific, repeatable job.

Using two photos: matching a background or style

As of 2026, the Gemini app lets you upload two images in one request and combine them, which opens up a genuinely useful trick for headshots: upload your selfie plus a photo of a background or lighting setup you like, and ask the model to put you in that setting rather than describing it from scratch in words.

"Using the lighting and background from the second photo, create a professional headshot of the person in the first photo. Keep their exact face and features unchanged." This works well for matching an existing team's headshot style, since you can use a colleague's existing photo (with their knowledge, obviously) as the style reference rather than trying to describe their background and lighting setup in prose. It's also useful for matching a specific office or location you've photographed yourself but don't want to physically go stand in front of.

This is one case where the manual route genuinely has an edge over most dedicated headshot tools, which typically only accept a single reference photo of the subject and choose the background from a preset list rather than an arbitrary second image.

Manual Nano Banana vs a dedicated headshot tool

	Manual (Gemini app)	FreeHeadshot.org
Model	Nano Banana (Gemini 2.5 Flash Image)	Same model
Prompt writing	You write and iterate it yourself	Pre-tuned per style, no prompt writing
Consistent multi-shot set	Manual, regenerate and compare yourself	One upload returns a matched set
Aspect ratio for LinkedIn/resume/etc	Manual crop after generation	Sized per platform automatically
Cost	Free (Gemini app free tier) or API cost	Free tier, then $9 to $49 one-time
Style catalog	Whatever you can describe in a prompt	100+ named, pre-built styles
Post-processing (skin texture, grain, watermark)	None, raw model output	Automated grade + texture pass

If you only need one image and don't mind iterating, the manual route costs nothing and works. If you want a matched set across multiple styles without writing and rewriting prompts, that's the gap a dedicated tool closes.

When manual prompting is genuinely the better choice

To be fair about it: if you want full creative control over an unusual request a preset style catalog won't cover, or you're comfortable iterating and only need one final image, going straight to Gemini is completely reasonable and free. A dedicated tool earns its keep specifically when you want a consistent set across several styles without doing the prompt engineering yourself, not because the manual route is broken.

FAQ

Is Nano Banana free to use?

Yes, through the free tier of the Gemini app. Higher-capability tiers (including Nano Banana Pro) require a paid Google AI subscription for higher usage limits and priority access.

What's the difference between Nano Banana and Nano Banana Pro?

Nano Banana is the nickname for Gemini 2.5 Flash Image (2025). Nano Banana Pro is the nickname for Gemini 3 Pro Image (shipped June 2026), which adds accurate in-image text rendering, native multi-turn editing, and up to 4K resolution. For a plain headshot without text in the image, the difference is minor.

Can Nano Banana make a professional headshot from just one photo?

Yes. It reads the uploaded photo directly as part of its input and doesn't require multiple photos or a training step, unlike older AI headshot approaches from 2022 to 2023.

Why does my Nano Banana headshot look different every time I regenerate it?

The model doesn't lock identity between separate generations the way a dedicated identity-preserving pipeline does. Each regeneration is a fresh interpretation of your reference photo, which is why the face can drift slightly attempt to attempt.

Is FreeHeadshot.org the same model as Nano Banana?

Yes. We run Gemini 2.5 Flash Image in production, the same model behind the free tier of the Gemini app, with prompts pre-tuned per style so you don't have to write or iterate on them yourself.

Why does my AI headshot look too smooth or airbrushed?

Most image models default toward smoother skin unless the prompt explicitly asks for texture preservation. Adding an instruction like "preserve visible pores, no airbrushing" measurably changes the output, see prompt 16 above for the exact wording we use.

Is it safe to upload a selfie to the Gemini app?

It's covered by Google's general Gemini app privacy terms, which are reasonable but not written specifically around face-photo sensitivity. If that matters to you, check what a dedicated tool specifically says about photo retention before uploading anywhere, covered in more detail above.

Why does Nano Banana take a long time to generate my headshot?

If your session defaults to a reasoning-heavy model tier, it can add unnecessary time to a simple request. Switching to the "Fast" tier in the model menu is usually faster for a straightforward headshot and produces comparable quality for this use case.