DEV Community: Joseph Dillon

I Tested Nano Banana AI for a Week. Honestly, I Didn’t Expect It to Be This Good

Joseph Dillon — Tue, 12 May 2026 09:04:02 +0000

I’ve tested a lot of AI image tools over the past year. Some looked impressive for a few minutes, then became frustrating once you tried using them in a real workflow. Others generated pretty images but failed badly at consistency, editing, or speed.

So when I saw people talking about Nano Banana AI on Artlist
, I honestly assumed it was another overhyped image model with a funny name.

I was wrong.

Nano Banana is based on Google’s Gemini 2.5 Flash Image technology and focuses heavily on fast image generation and conversational editing.

After spending several days testing it for thumbnails, social posts, image editing, and character consistency, I finally understood why creators were suddenly talking about it everywhere.

The First Thing That Surprised Me Was the Speed

Most AI image tools interrupt your creative flow.

You write a prompt, wait, tweak it, wait again, regenerate, then repeat until you lose patience. Nano Banana felt different almost immediately because the outputs generated fast enough that I could actually iterate naturally.

That speed matters more than people think.

When you’re working on thumbnails, Instagram posts, ad creatives, or short-form content, waiting two or three minutes per generation completely breaks momentum.

Nano Banana is specifically designed for low-latency workflows and rapid iteration.

I started by testing YouTube thumbnail concepts.

Normally, AI tools struggle here:

inconsistent faces
weird lighting
drifting compositions
distorted text
different-looking characters every generation

Nano Banana handled this surprisingly well.

I uploaded a rough thumbnail reference and asked it:

“Keep the same pose and expression, but make the lighting cinematic and add a darker tech background.”

The result looked more like a Photoshop edit than a full regeneration.

That was the first moment where I realized this model was built around workflows instead of random image generation.

Editing Images With Plain English Feels Weirdly Good

This is probably Nano Banana’s strongest feature.

Most AI image models are great at generating new images but terrible at editing existing ones precisely. Once you want to change a small detail, things start breaking.

Nano Banana approaches editing differently.

You can literally type:

“Change the hoodie to black”
“Move the product slightly left”
“Turn this into golden hour lighting”
“Blur the background more”
“Make this look like an Instagram ad”

And most of the time, it understands exactly what you mean.

Artlist describes Nano Banana as an editing-focused AI model that preserves scene consistency while applying targeted edits, and after testing it myself, that description feels accurate.

The biggest difference is that the image doesn’t completely collapse every time you edit something.

The face remains recognizable.
The composition stays stable.
The lighting direction still makes sense.

That sounds basic, but honestly… many AI image models still fail badly at this.

Character Consistency Is Better Than Most AI Tools

This is where many image models completely fall apart.

Generate the same person twice and suddenly:

the jawline changes
the hairstyle shifts
clothing becomes different
proportions look wrong

Nano Banana handled consistency much better than I expected.

I tested a fictional creator persona across multiple scenes:

coffee shop setup
studio desk setup
podcast room
outdoor lifestyle shots
product holding poses

The same person remained recognizable across most generations.

Not perfect every single time, obviously. But consistent enough that it actually felt usable for branded content or storytelling projects.

Google’s Gemini image models are heavily focused on subject consistency and conversational editing, which becomes obvious once you test multiple iterative generations.

I can already see why creators are using it for:

AI influencers
UGC-style content
campaign assets
ad creatives
content series

Multi-Image Prompting Was Surprisingly Useful

One feature I didn’t expect to use much was multi-image prompting.

Nano Banana AI allows multiple reference images during generation workflows.

I uploaded:

a product image
a lighting reference
a background mood reference
a pose reference

Then I asked the model to combine everything into a single composition.

Normally this creates chaotic results in many AI tools.

Nano Banana handled it surprisingly well.

The final outputs looked coherent instead of stitched together awkwardly.

For content creators, agencies, and e-commerce brands, this is actually huge because it reduces the amount of manual compositing work dramatically.

The Most Impressive Part Was Contextual Editing

This was probably my favorite test.

I uploaded a normal indoor portrait and asked:

“Turn this into a cyberpunk nighttime scene while keeping the same facial expression and camera angle.”

Most AI tools would regenerate the entire image from scratch.

Nano Banana preserved the original structure while transforming the environment around it.

That changes the editing experience completely.

It feels less like:
“Generate another image.”

And more like:
“Collaborate with the current image.”

That’s a subtle difference, but after using many AI image tools, it stands out immediately.

It’s Not Perfect, But It Feels Workflow-Ready

Of course, it still has flaws.

Hands occasionally become strange.
Typography can still break sometimes.
Complex prompts can confuse the model if overloaded.

But compared to many AI image tools I’ve tested recently, Nano Banana feels closer to something creators would actually use daily instead of occasionally.

Even major tech publications have started discussing how Gemini Flash Image editing is pressuring traditional editing software because of its precision and consistency.

What surprised me most wasn’t the image quality itself.

It was the practicality.

Nano Banana feels designed for creators who need:

fast iterations
realistic edits
thumbnails
social media visuals
product shots
campaign assets
consistent characters

Not endless prompt engineering experiments.

After testing it properly, I’d honestly describe Nano Banana as one of the first AI image tools that feels workflow-ready instead of demo-ready.

The name still sounds ridiculous though.

References:
https://x.com/testingcatalog/status/2053924078537318681?lang=en
https://artlist.io/ai/models/nano-banana-2
https://www.businessinsider.com/google-gemini-flash-nano-banana-viral-image-editor-adobe-photoshop-2025-8

How I Tested a Text to Video Tool in a Real Workflow

Joseph Dillon — Mon, 15 Dec 2025 10:49:31 +0000

Text to video tools sound exciting on paper. Type a prompt, get a video, move on. In reality, most developers and product teams want to know one thing. Does it actually help in real work?

I decided to test a text to video AI tool inside an actual workflow. Not a demo. Not a one off experiment. A real use case with deadlines, revisions, and feedback.

This post shares what worked, what did not, and where this type of tool fits today.

Why I Tried Text to Video AI

I often need short videos. Product demos, landing page previews, onboarding clips, and quick explainers for internal teams. Traditional video creation takes time. Scripts, screen recordings, edits, exports. It adds up fast.

I wanted something that could help me:

• Create fast visual drafts
• Test ideas before committing to production
• Support non designers on the team
• Reduce back and forth during early stages

That is where text to video AI looked promising.

The Workflow I Used

I kept the setup simple and close to how most teams work.

Step one was writing a rough script. Nothing polished. Just clear sentences explaining a feature or flow.

Step two was generating short video clips from those prompts. I tested different tones. Product focused. Neutral. Slightly creative.

Step three was placing the output into real contexts. A landing page draft. A product walkthrough. An internal demo deck.

This helped me judge the tool based on usefulness, not novelty.

What Worked Well

The biggest win was speed. I could turn an idea into a visual in minutes. That alone made it useful during early planning.

Another strong point was clarity. The videos helped explain concepts that were hard to describe with text alone. This was helpful for async communication and early stakeholder reviews.

I also noticed that the tool worked best when prompts were clear and structured. Simple language produced better results than vague descriptions.

During this test, I explored a few platforms, including this text to video option: Kling 2.5 Turbo. It handled short, focused prompts well and fit naturally into quick iteration cycles.

Where It Fell Short

Text to video AI is not a replacement for real video production. At least not yet.

Fine control is limited. You cannot easily tweak small details the way you would in a video editor. If something feels slightly off, you often need to regenerate instead of adjusting.

Consistency can also be a challenge. When you need multiple clips that look and feel the same, it takes effort to guide the tool with careful prompts.

This means the output works best as a draft or supporting asset, not a final polished video.

How It Fit the Team

This tool was most useful for:

• Early stage demos
• Internal presentations
• Product concept previews
• Quick onboarding explanations

It helped non technical teammates understand features faster. It also reduced the pressure on designers and video editors during early phases.

Once the direction was clear, we still moved to traditional tools for final assets.

Tips If You Want to Try It

Based on this test, here are a few practical tips.

Start with short videos. Thirty to sixty seconds works best.

Write prompts like instructions, not marketing copy.

Test videos inside real layouts. Context matters.

Use it early. Do not wait until the final stage.

Treat the output as a draft, not a finished product.

You Should Try it For Real Results

Text to video AI is most useful when you treat it as a thinking tool, not a shortcut to final content. It helps you explore ideas, explain flows, and move faster during planning.

For developers and product teams, that can be enough to justify using it. Not because it replaces anything, but because it helps you decide what to build next with more clarity.

If you are curious, try it inside a real workflow. That is where its strengths and limits become clear.

References:
https://dev.to/alifar/automating-text-to-video-pipelines-with-sora-2-and-n8n-lh0
https://workspace.google.com/resources/text-to-video/

A Simple Guide to Stock Photos for UI and UX Work

Joseph Dillon — Thu, 11 Dec 2025 07:08:54 +0000

Good visuals shape how users feel the moment they land on a screen. A single photo can set the mood, support your message, and help people understand the purpose of your design. Whether you build a website, a mobile app, or a quick prototype, the images you choose affect the entire experience.

As someone who works on UI and UX projects every week, I spend a lot of time searching for photos. I use them in layouts, case studies, landing pages, and social posts. Strong visuals save time and give your work a consistent style.

This guide shows you the best places to find stock photos that fit modern digital design. Each site here offers clean, useful images that help your UI feel polished and focused.

Free Stock Photo Platforms

You have many places to find clean, modern visuals for UI and UX work. These sites help you move fast without losing quality.

Pexels and Pixabay

Both are strong choices when you need quick results.
• Pexels gives you sharp photos with simple tags and fast search.
• Pixabay offers a wide library of photos and videos with easy filters.
Both are good for common UI scenes like workspaces, devices, and lifestyle shots.

Unsplash

A favorite for designers. You get natural lighting, real life scenes, and a huge range of categories. The photos work well in hero sections, mockups, and onboarding screens.

Motion Array

If you need a wider mix of creative styles, check this collection of Stock photos. It includes modern visuals that fit digital products, landing pages, and design systems. You can browse by style, tone, and subject, which helps when you want consistency across your UI.

FOCA

A small but beautiful library with calm, minimal photos. Great for dashboards, portfolios, and clean UI layouts.

FreePhotos.cc

Good for quick browsing.
• Clear preview
• More than 100 categories
• Simple navigation
Useful when you want photos that match a specific theme.

Magdeleine

A curated site with warm, detailed photography. You can filter by dominant color, which helps when you design screens with a strong palette.

Tools for Color, Style, and Mood

Your images should support the mood of your product, not fight against it. These tools help you stay aligned with your visual direction.

Duotone

A simple tool that applies a controlled two color filter to any photo. Perfect for hero banners, minimal landing pages, or design systems with strict color rules.

pxhere

Useful when you want color filtering. You can quickly match visuals to a palette and keep your UI style consistent.

Reshot

Great when you want photos that feel natural instead of staged. Ideal for apps that need real user moments.

Photos for Inclusive and Realistic Design

Your users come from different backgrounds. Your visuals should reflect that.

WOCinTech Chat

A library featuring women in tech from diverse communities. Strong for product pages, case studies, and feature sections.

The Gender Spectrum Collection

Images of trans and non binary models in realistic settings. Helpful when you design for inclusive products.

Niche Photo Resources

Some projects need visuals that feel specific. These sites help when you want a certain theme or style.

FoodiesFeed

A strong choice for food apps, restaurant pages, and blog layouts. Photos look natural, sharp, and clean. You get overhead shots, close ups, and full table setups.

New Old Stock

This site collects vintage photos from public archives. Great for timelines, history pages, and branding projects that want a classic tone.

SkitterPhoto

A mix of everyday scenes, nature, and objects. You can sort by color or check the newest uploads. It helps when you need something simple and neutral.

StockSnap.io

A large library that grows every week. You can search by mood or keyword. Many UI designers use it for backgrounds, hero sections, and mood boards.

How to Choose the Right Photo for UI and UX Work

The right photo supports your design. It guides the user’s attention and helps them understand the purpose of the screen. These simple points keep your visuals clear and effective.

Match the photo to the user’s goal

Before you pick anything, decide what the user should feel. Calm, focused, inspired, curious. Choose an image that fits that state.

Keep distractions low

Avoid photos with heavy detail when placing text on top. Simple backgrounds make your content readable.

Check the color balance

Your UI palette should still lead the experience. Pick photos that blend with your colors, not fight against them. Tools like Duotone or the library at Stock photos help you keep everything aligned.

Stay consistent across screens

If your hero image is soft and minimal, follow that tone in your inner pages. A mixed visual style makes the UI feel scattered.

Test in real layouts

Always drop photos into your Figma or design file. A photo that looks nice alone can behave differently once it sits inside a real interface.

If you liked the post, thumbs up and comment below what’s your favourite platform for selecting photos for your UX/UI projects.

References:
https://medium.com/design-bootcamp/7-free-stock-image-websites-for-every-ui-ux-designer-e48f0277f985
https://elements.envato.com/learn/stock-photography-guide-designers
https://www.istockphoto.com/photos/ux-ui-design

Reliable API Saved My Time

Joseph Dillon — Wed, 10 Dec 2025 17:28:55 +0000

I joined this project with a clear goal. Build API tools that help teams ship faster and keep their workflows simple. My role at Qwilr was to understand how people work, then shape the product so it fits their daily needs. I spent time with customer success, engineering, and sales to learn where users struggle. These conversations helped me map what the API should do and how it should feel.

I wrote clear product requirements and turned complex ideas into simple steps for the team. I defined the core use cases, from automated document creation to seamless integration with CRM and billing tools. I worked closely with engineers to make sure the API stayed easy to implement. We focused on clean endpoints, stable performance, and predictable behavior.

I also built sample projects to show how the API works in real situations. These examples helped developers save time and understand the value faster. Along the way, I created simple guides, short tutorials, and onboarding flows that removed friction for new users.

The final result was a reliable API that teams could trust. It supported their workflow, reduced manual work, and opened new ways to scale their processes.