How to Generate Nano Banana Images with Templates ?

If you just want your first result fast:

Open the Discover
Search for Nano Banana
Pick a template → choose an aspect ratio (recommend starting with 3:4 for portraits and 9:16 for vertical short-form)
Upload a reference image only if the template asks for it
Generate → download as PNG

This guide shows you the reliable way to get clean Nano Banana results: how to choose the right template, which aspect ratio to use, how to iterate without breaking consistency, and how to troubleshoot the common failure modes.

If you’re here for portrait-style results (headshots, photoshoots, profile pictures)—you’re not alone. A large portion of the template library is portrait-focused, and this is also the easiest place to get “premium-looking” output if you pick the right template and feed it a clean reference image when needed.

What are Nano Banana and Nano Banana Pro?

First, a quick clarification: Nano Banana / Nano Banana Pro aren’t standalone apps. They’re nicknames (aliases/codenames) Google uses inside the Gemini ecosystem for two generations of AI image generation + editing models.

Nano Banana: maps to Gemini 2.5 Flash Image — optimized for speed and high-frequency creation. Great for fast iterations, compositing images, local edits, and keeping character consistency.
Nano Banana Pro: maps to Gemini 3 Pro Image — optimized for higher fidelity and stronger control. It’s positioned for production-grade deliverables, clearer text rendering (posters/diagrams), finer lighting/camera control, and stronger consistency/control (enterprise setups can support more reference images).

Sources (official): Google DeepMind / Gemini / Google Cloud.

Practical way to think about it in Artifox:

Use Nano Banana to explore ideas quickly (speed-first)
Switch to Nano Banana Pro for the final deliverable (quality + control-first)

Why templates beat writing prompts from scratch

You can write prompts manually, but templates win when you care about speed and consistency:

Less guesswork: composition, lighting, materials, and camera language are already tuned
More consistent results: easier to keep a recognizable style across a series
Faster iteration: change one input, keep everything else stable
Trend alignment: popular prompt patterns are frequently packaged into reusable templates

The most common template use case: portraits / headshots / photoshoots

Beyond avatars and covers, portrait templates (Portrait / Headshot / Photoshoot / Studio) are typically the highest-usage category because they’re sensitive to lighting, skin texture, lens look, and background blur. Hand-writing prompts for portraits is where most people hit the “uncanny / plastic skin” wall.

Use this quick guide:

Portrait photoshoot / professional headshot: look for templates that mention portrait, headshot, photoshoot, studio
Need it to look like a specific person: prefer templates that require a reference image
Want clean “camera-like” depth of field: choose templates with natural background blur in the previews

Start here: use the right entry point

Discove Templates

Template entry point

If you’re creating more than one image, log in first so you can favorite templates and reuse your history.

Generate Nano Banana images with templates (5 steps)

Find the right Nano Banana template

Search the library with:

Nano Banana
Nano Banana Pro
banana

If you’re generating portraits/headshots, also search:

portrait
headshot
photoshoot
studio

Don’t judge by the title alone. Use previews to check:

The texture/style (realistic vs cartoony vs 3D)
Whether it nails expressions (often the difference between “meh” and “click”)
Background suitability (clean studio vs scene-based)

For portraits specifically, I also check:

Skin texture: does it look natural (not waxy/plastic)?
Face stability: eyes/teeth/hairline look consistent
Lens look & blur: does the depth-of-field feel like a real camera?

Search and pick a Nano Banana template

Choose the aspect ratio based on your use case

Supported ratios: 1:1, 9:16, 16:9, 3:4, 4:3, 2:3, 3:2.

If you’re unsure, use this:

Avatars / stickers / product images: 1:1
Short-form (Reels/TikTok/YouTube Shorts) (recommended): 9:16
YouTube thumbnails / wide banners: 16:9
Portrait / headshot (recommended): 3:4
Photo-like editorial images: 3:4

Across common social video sizing guides, 9:16 (often 1080×1920) is the standard recommendation for Shorts/Reels/TikTok (for example, Kapwing and HopperHQ both highlight 9:16 as the most common vertical format).

A lot of “bad model” results are actually “wrong ratio” results. If the subject looks tiny or cramped, change the aspect ratio first.

Check if the template requires a reference image

There are two common template types:

No reference image needed: text-only generation
Reference image required: better for matching a specific subject/composition

For portraits, here’s the simple rule:

If you want it to look like you / a specific person, use a reference-required template.
If you only want a photoshoot vibe (lighting + lens look) and don’t care about identity, start with a non-reference template.

If the template asks for an upload, keep your reference image:

Clear, not over-compressed
Not heavily over/under-exposed
With a clean background (unless the background matters)

Upload a reference image

First generation: keep inputs minimal

For your first run, use a short description. Avoid stacking adjectives.

Examples (copy/paste and adjust):

“A tiny banana character, smiling, clean light background”
“Mini banana mascot holding a coffee cup, studio lighting, shallow depth of field”

If you’re generating portrait-style images, start with one of these (keep it short):

“Portrait headshot, studio lighting, clean background, shallow depth of field, natural skin texture”
“Professional photoshoot portrait, soft light, realistic camera look, background blur”

Once you like the baseline, iterate on one dimension:

expression (happy / surprised / angry)
prop (coffee / book / mic)
background (solid / room / street)

Prompt + generate button

Iterate like a pro: change one thing at a time

If you want a consistent “series,” the rule is simple: change one variable per iteration.

For example, for a 6-image set:

Keep: aspect ratio, lighting, background style
Change: expression / prop / pose

This is how you get a clean, recognizable Nano Banana look without randomness taking over.

Iteration comparison

3 reliable setups (steal these)

Setup A: avatar / sticker (1:1)

Ratio: 1:1
Focus: clear subject + obvious expression
Background: solid or subtle gradient

Setup B: short-form cover (9:16)

Ratio: 9:16
Composition: subject centered slightly high, leave space below
Background: higher contrast for feed visibility

Setup C: wide banner / thumbnail (16:9)

Ratio: 16:9
Composition: subject left/right, leave headline space
Background: keep it clean (avoid tiny clutter)

Setup D: portrait photoshoot / headshot (3:4 / 2:3)

Ratio: 3:4 (editorial / photoshoot and easier for mobile-first portrait crops)
Composition: start with headshot or half-body (full-body is less stable)
Lighting: studio/soft light templates tend to look more premium
If identity matters: use a reference-required template instead of trying to “prompt” a face

Export format: PNG

Outputs are currently PNG.

Batch generation isn’t available right now. If you’re building a set, track your template + ratio + baseline text, then iterate by changing one thing at a time.

Troubleshooting checklist

The subject is too small / too much empty space

Change ratio (try 3:4)
Choose a closer, portrait-style template

The background is busy and cheap-looking

Use clean studio/solid background templates
Reduce background detail in your text

Reference upload: “closer, but weird”

Use a cleaner reference image (lighting matters)
Crop to isolate the subject

Portraits: uncanny face / plastic skin / weird eyes

Switch to a dedicated portrait/headshot/photoshoot template (don’t force a cartoon-style template)
Prefer reference-required templates and use a clean, front-facing reference
Keep the first prompt minimal (lighting + lens look), avoid stacking facial-detail adjectives
If eyes/teeth/skin look off, it’s usually reference quality or template mismatch—swap the reference first, then switch templates

Frequently Asked Questions

What is the difference between Nano Banana / Nano Banana Pro?

In plain terms: Nano Banana is for fast iteration; Nano Banana Pro is for higher-fidelity, tighter control, and production-ready assets.

Dimension	Nano Banana	Nano Banana Pro
Underlying model	Gemini 2.5 Flash Image	Gemini 3 Pro Image
Positioning	Fast / casual creation	Pro-grade quality + control
Text rendering (posters/diagrams)	OK	Clearer and more reliable
Control (lighting/camera/composition)	Baseline	Finer control
Reference images	Supported	Enterprise setups can support more references for brand/character consistency

Quick selection rule:

For portrait/photoshoot vibes and rapid exploration: start with Nano Banana
For brand assets, posters with readable text, diagrams, or picky lighting/camera requirements: use Nano Banana Pro

Can I use templates without writing prompts?

Yes. Templates are designed to remove prompt complexity. A single, clear sentence is usually enough to get a usable first result.

Which aspect ratios are supported?

1:1, 9:16, 16:9, 3:4, 4:3, 2:3, 3:2.

If you only want the two safest picks:

Portraits / headshots: 3:4
Vertical short-form covers (Reels/TikTok/Shorts): 9:16

Why do some templates require a reference image?

Those templates are built to match a specific subject/composition. Uploading a reference makes results more aligned with your intent.

Do I need a reference image for portraits?

Not always.

If you want a photoshoot vibe and don’t care about identity: a non-reference portrait template can be enough.
If you want it to look like you: use a reference-required template for much better stability.

What file format do I get?

PNG.

Generate your first Nano Banana image in minutes

Browse templates while logged out, then log in to generate with one click.

Open the template library

Key Takeaways

Templates are the fastest, most consistent way to get Nano Banana results
Discove Templates
Choose ratio by use case: 1:1 for avatars, 9:16 for short-form, 16:9 for wide covers
Keep the first prompt short; iterate after you get a clean baseline
For consistency, change one thing at a time