This AI Creates Images That Sell

The End of Generic B2B Outreach

Cold B2B outreach has a spam problem. Sales teams blast out thousands of near-identical emails, and prospects respond with the only rational move: ignore almost all of them. Even “personalized” messages that swap in a first name or company field rarely crack single-digit reply rates.

Hyper-personalization promises better numbers, but humans cannot handcraft 500 bespoke pitches a day. That gap is where Nano Banana and tools like n8n get interesting. Instead of tweaking copy, they generate a completely unique image for every prospect that visually centers the prospect’s own brand.

Picture opening your inbox and seeing a thumbnail of a hyperrealistic phone on a desk, notification badges lighting up, your company’s logo rendered perfectly on-screen in 4K. Nano Banana’s image-to-image model ingests a logo and business name, then weaves them into a predesigned scene with high-fidelity text and branding. No watermarks, roughly $0.10 per generation, and output tuned for 16:9 social and email previews.

Psychologically, that hits very differently from a generic stock photo. Humans exhibit a strong self-referential bias: we pay more attention to things that mirror our identity, including names and symbols. A custom visual that literally showcases your logo and brand colors signals effort, relevance, and intent in a way a mail-merge line never will.

Sales teams already know that genuinely believable personalization can roughly double response rates compared with templated outreach. Now imagine pairing that uplift with an image that looks like a designer spent an hour mocking up a campaign just for you. The pitch stops feeling like a cold email and starts to resemble a preview of a partnership.

Under the hood, automation platforms such as n8n tie this all together for hundreds or thousands of contacts. A workflow can:

• Scrape leads and company data
• Enrich records with emails and logos
• Generate a custom Nano Banana prompt per prospect
• Produce and store unique images in Google Drive or Airtable

That pipeline runs on autopilot. The result: every outbound sequence can carry a one-of-a-kind, on-brand visual hook tailored to each prospect, at the scale of a traditional email blast.

Your New Automation Super-Stack

Sales teams chasing higher reply rates do not need another point solution; they need a coordinated automation stack. The workflow here chains together scrapers, enrichment tools, prompt engines, and storage so a single lead can go from URL to bespoke, on‑brand image in under a minute.

At the center sits n8n, the visual automation platform that wires everything together. Each service becomes a node: scrape a site, enrich the contact, generate a prompt, call an image model, then ship the finished asset to storage and your CRM—no manual copy‑paste, no brittle glue code.

For the visuals, Nano Banana handles the heavy lifting via the Kia API. You send a logo, brand name, and a structured prompt; Nano Banana returns a 4K, 16:9 image with clean text rendering and on‑brand composition, typically for around $0.10 per generation. n8n polls the Kia API until the state flips to “completed,” then pulls down the final image.

Lead generation starts upstream with Apify and Hunter.io. Apify crawlers collect hundreds of target companies in a niche—domains, LinkedIn profiles, job titles—at scale. Hunter.io then enriches those records with verified email addresses and confidence scores so you are not blasting guesses into the void.

Prompt quality decides whether your outreach looks magical or gimmicky, which is where ChatGPT comes in. n8n feeds company data, industry, and persona into ChatGPT to output a JSON prompt that describes a believable scenario for each prospect’s brand—no two prompts the same. Airtable stores all of this: lead rows, prompt text, image URLs, send status, and performance metrics.

Because Nano Banana and email clients both prefer URLs, the system uses ImageBB as a temporary image host. n8n uploads the generated asset to ImageBB, grabs a hotlink-safe URL, and injects it into the email template as an inline image. Once campaigns finish or assets need long‑term retention, n8n mirrors everything into Google Drive, organized by campaign, segment, or account tier.

Stitched together, this stack behaves like a dedicated creative ops team that never sleeps: constantly scraping, enriching, generating, and filing away assets while your sales reps focus on actual conversations.

The Basic Recipe: One Image, Endless Possibilities

Sales teams start with a single, controlled workflow: one image in, one image out. No scraping, no agents, just a clean path from logo to finished creative so you can see exactly how Nano Banana behaves before you scale it to hundreds of prospects.

First move is manual. You open n8n, hit execute, and pass two inputs: a logo file and the business name. That could be a PNG pulled from your desktop and “Acme Analytics” typed into a field, or any brand asset you want Nano Banana to remix.

Because Nano Banana expects a public image URL, n8n immediately ships that logo to ImageBB. ImageBB hosts the file and returns a direct, hotlinkable URL that Nano Banana’s backend can fetch reliably. No need to expose your own storage bucket or worry about expiring share links.

With a URL in hand, n8n calls the Nano Banana API through a generic HTTP node (Jack Roberts uses Kia API as the bridge). The request includes: - The ImageBB URL - The business name - A hardcoded prompt defining the visual style - Output settings like resolution and aspect ratio

In Roberts’ demo, the prompt asks for a hyperrealistic phone-on-desk scene, 4K, 16:9, with notifications branded to the target company. Hardcoding the prompt guarantees every lead gets the same aesthetic while still swapping in unique logos and names, which keeps the output on-brand and predictable.

Nano Banana responds asynchronously, so the workflow polls the job status until the API reports “completed.” That loop prevents half-baked results and lets you see failures when you run out of credits or pass a bad URL. Each generation costs roughly $0.10, cheap enough to experiment aggressively without design-team bottlenecks.

Once the image finishes rendering, n8n grabs the final, watermark-free file and uploads it straight to Google Drive. You end up with a neatly organized folder of personalized creatives, one per run, ready to drop into outbound sequences or landing pages.

Similar n8n recipes already power production flows, like Generate AI Product Photos using Gemini Nano Banana with Jotform and Google Sheets, proving this single-image pattern scales cleanly to more complex pipelines.

Connecting the Dots: Your First n8n Workflow

Open n8n and you land on a blank canvas that behaves more like a circuit board than a flowchart. Each node is a component: trigger, HTTP call, loop, and storage, all wired left to right. For this first build, you only need five nodes to go from raw idea to a finished, saved image.

Start with a Manual Trigger node. It acts as your on/off button during development, letting you fire the workflow on demand instead of waiting for webhooks or schedules. You can pass test data into it—like a logo URL, brand name, or prompt—so every run mimics a real lead.

Next comes the HTTP Request node that talks to ImageBB. Configure it as a POST request to the ImageBB upload endpoint, drop in your API key as a query parameter, and send the binary image from your input. ImageBB returns a permanent, CDN-backed URL, which becomes the reference image Nano Banana needs.

Add a second HTTP Request node for the Kia API, which wraps the Nano Banana model. Point it at the kia.ai endpoint, choose JSON as the body type, and include fields for the ImageBB URL, your prompt template, and any style or resolution settings. Jack Roberts hardcodes a “hyperrealistic phone on a desk” prompt, then dynamically injects the brand name and logo reference.

Kia’s response does not hand you a finished image; it gives you a job ID and a status like “pending” or “running.” That asynchronous design means your workflow must poll for completion instead of assuming instant results. Without a loop, you either miss the asset or stall the whole run.

Drop in a Wait or loop pattern using n8n’s built-in nodes. A common setup hits a “Get job status” HTTP Request every 5–10 seconds, checks the returned JSON field `state`, and continues looping while the value is not `completed`. You can cap retries, log failures, and branch to an error path if the job times out or returns “failed.”

Once `state` flips to `completed`, the API includes a final image URL or file blob. Feed that into a Google Drive node configured in “Upload” mode, choose a target folder like `/Nano-Banana-Campaigns`, and set the filename to something traceable—company name, timestamp, campaign ID. Now every generated image lands in Drive, versioned, searchable, and ready for your outbound sequence.

Why This Changes B2B Sales Forever

Crowded B2B inboxes reward whatever breaks the pattern. A cold email that opens with a wall of text blends into the noise; an email that leads with a bespoke, on-brand image hits the brain differently. That visual pattern interrupt buys the extra three seconds every salesperson is fighting for.

Most “personalized” outreach today means swapping {{FirstName}} and maybe {{Company}} into a template. Studies from platforms like HubSpot and Campaign Monitor routinely show that genuinely personalized emails can 2x response rates over generic blasts. Dropping a prospect’s actual logo into a tailored visual narrative pushes that personalization from cosmetic to unmistakably real.

Nano Banana plus n8n turns this into a repeatable system, not a one-off stunt. You feed in a logo, company name, and a prompt; the workflow returns a 4K, 16:9, hyperreal image for roughly $0.10 per generation, with no watermark. At scale, that makes “designer-level” personalization economically viable for hundreds or thousands of prospects per campaign.

The creative surface area is huge. Picture a prospect’s logo floating in a cereal bowl with the caption, “We eat competitors for breakfast,” sitting above a one-line pitch. Or their logo glowing on a phone lock screen, surrounded by notification badges that mirror the exact KPIs your product claims to improve.

This kind of image does three jobs instantly: - Signals you did real homework - Demonstrates your product story in one glance - Creates an asset the prospect can forward internally

Because Nano Banana handles image-to-image and logo placement programmatically, sales teams can tie it directly into lead scraping, enrichment, and sequencing. A workflow that once lived in Figma and manual copy-paste now lives in APIs and triggers. That shift moves personalization from “nice-to-have craft project” to an always-on growth engine for B2B teams.

Building the Full Lead-Gen Engine

Generic one-off image tests are cute; a full lead-gen engine is brutal. Scaling this idea means wiring Nano Banana and n8n into a pipeline that never stops feeding your CRM with fresh, personalized visuals aimed at real decision-makers, not hypothetical personas.

At the highest level, the advanced workflow runs as a tight five-step loop: - Scrape: pull hundreds or thousands of companies in a niche - Enrich: attach domains, roles, and verified email addresses - Prompt: generate JSON-structured, per-lead prompts - Generate: call Nano Banana’s API for custom images - Distribute: push assets into outbound tools and storage

Scraping replaces manual prospect hunting. Instead of you Googling “SaaS marketing agencies London” and copying domains into a spreadsheet, n8n hits APIs, SERPs, or databases, then dumps structured company rows into Airtable along with logos and basic metadata.

Enrichment turns those rows into people. You bolt on services that append job titles, LinkedIn URLs, and direct emails, so each record in Airtable now represents an actual VP of Sales or Head of Marketing, not just a faceless brand.

Prompting is where the personalization gets sharp. An LLM reads each enriched record and outputs a JSON prompt that bakes in company name, role, pain points, and visual style, ready for Nano Banana to turn into a 4K, 16:9, logo-infused hero image.

Generation and distribution then run on autopilot. n8n uploads logos to ImageBB, calls Nano Banana via the Kia API, polls until completion, stores results in Google Drive or Airtable, and syncs image URLs into your outreach stack. For broader context on image APIs and tooling, Google’s Google AI Studio Documentation shows how similar pipelines slot into modern workflows.

Compared to the original single-image demo, this system behaves like infrastructure. Once configured, it quietly converts every new lead into a tailored visual asset, 24/7, without you lifting a finger.

Finding Your Perfect Customers with Apify

Cold outreach lives or dies on targeting, and that starts with a clean, high-intent lead list. Jack Roberts leans on Apify as the data firehose, turning a vague ICP like “online business coaches” into hundreds of concrete profiles your automation can actually use.

Apify runs on “actors” — prebuilt scraping bots — and the star here is Leads Finder. Instead of wrestling with custom scrapers, you open Leads Finder, plug in your criteria, and Apify fans out across the web to assemble a prospect list from public sources.

Configuration happens in a single JSON-style form. You define who you want with filters for: - Industry / niche (e.g., “business coach”, “fitness coach”, “SaaS consultant”) - Job titles (e.g., “Founder”, “Coach”, “Marketing Consultant”) - Geography (e.g., “United Kingdom”, or narrower regions and cities)

Type “coaches in the UK” as your seed, then tighten it. You might combine “business coach” as a keyword, “United Kingdom” as location, and “Founder OR Coach” as title. Leads Finder then crawls business sites, directories, and social profiles to pull back only companies and people that match those constraints.

Each run returns structured records, not a messy HTML dump. A typical row includes: - Company name and domain - Website URL and description - Contact names and roles - LinkedIn profile URLs - Social links and sometimes phone numbers

You control scale with run limits. Want a small test batch? Cap it at 50–100 results. Ready to feed your full Nano Banana engine? Push it to 500+ and let Apify chew through the query while n8n waits for the payload.

When the run finishes, Apify exposes everything as a dataset. From there, you hit “Export” and download a CSV that slots neatly into the next stage: enrichment, prompt generation, and image creation. Each row becomes a unit of work for your workflow — one company, one contact, one hyper-specific Nano Banana image, and one email that doesn’t feel like spam.

Reverse-Engineering Viral Prompts with ChatGPT

Generic prompts die quickly in automation. You cannot scale a campaign on “a nice image of a phone on a desk” and expect Nano Banana to keep spitting out usable, on-brand visuals for hundreds of different companies. You need a dynamic prompt template that behaves more like code than copy.

Start by stealing with your eyes. Scroll X.com for high-performing B2B visuals: SaaS dashboards bursting out of laptops, hyperreal phones with notification stacks, or founder photos composited into dramatic product scenes. When a concept feels “this would stop a CMO mid-scroll,” save the image and feed it directly into ChatGPT‑4.

The trick is not asking, “How do I recreate this?” but “Reverse‑engineer this as a JSON prompt.” ChatGPT can deconstruct the composition into structured pieces: camera angle, subject layout, depth of field, color palette, lighting, and text treatment. You move from vibes to a machine-readable spec that Nano Banana can hit over and over.

A good JSON prompt looks like a config file, not a paragraph. Instead of one long sentence, you get a schema with keys such as `"subject_style"`, `"background_style"`, `"lighting"`, `"text_overlay"`, and `"aspect_ratio"`. You can even mirror Nano Banana’s own parameters: `"resolution": "3840x2160"`, `"ratio": "16:9"`, `"quality": "high"`.

This is where personalization sneaks in. Instruct ChatGPT‑4 to expose every sales-relevant detail as a variable: `{brand_name}`, `{logo_text}`, `{industry}`, `{pain_point}`, `{cta_phrase}`, `{primary_color}`, `{dashboard_metric}`. Your JSON becomes a template that n8n can fill row by row from your lead list.

A typical structure might look like:

• `"prompt_core": "hyperrealistic image of a smartphone on a wooden desk with notifications"`
• `"brand_integration": "{brand_name} logo on screen and lockscreen"`
• `"copy_overlay": "{logo_text} · {industry} leads · {dashboard_metric}"`

Because the JSON stays consistent, Nano Banana receives the same skeleton for every generation. Only the variables change, so the system outputs a coherent visual series instead of a random collage of styles. That consistency matters when you send 500+ variations in a single campaign.

Structured prompts also make debugging trivial. If 20 images come back with unreadable text, you know to tweak `"text_style"` or `"font_weight"`, not rewrite the whole prompt. Over time, your JSON template becomes an asset: a battle‑tested “visual macro” that reliably turns scraped data into sales‑ready images at roughly $0.10 a pop.

The Final Assembly Line: Advanced n8n Logic

Automation stops being a parlor trick once you wire every piece into a single, relentless assembly line. In n8n, that line starts with your source of truth: an Airtable or Google Sheets node that pulls in the scraped leads, one row per company, complete with website, industry, and contact data. That sheet effectively becomes your production queue for outbound creativity.

From there, a Split In Batches node or a simple loop takes over, processing leads one at a time instead of slamming every API at once. You can throttle batches to 5, 10, or 20 records to stay under Nano Banana and OpenAI rate limits while keeping throughput high. Each iteration carries a single lead’s data payload down the chain.

Next comes the brains of the operation: a ChatGPT node that injects that lead data into your JSON prompt template. Company name, niche, and value prop flow into the same structured schema you engineered earlier for Nano Banana. The result is a per-lead JSON block that keeps style consistent while making the content feel eerily specific.

n8n hands that JSON to your image-generation sub-workflow, the same Nano Banana pipeline you built in the basic recipe. Instead of a static prompt, the sub-flow now receives a fully parameterized prompt with the company name, logo URL, and any campaign angle you want to test. One call in, one custom image out, at roughly $0.10 per generation.

You can trigger this sub-flow with a Call Workflow node, passing only the fields Nano Banana needs: prompt JSON, logo URL, and maybe a target resolution like 4K 16:9. The sub-flow handles the ImageBB upload, Nano Banana API call, and polling loop until the state flips to “completed.” It returns a clean image URL or file ID back to the parent workflow.

Once n8n has that URL, the line closes the loop by updating the original Airtable or Sheets row. A single Update Record or Update Row node writes the image URL into a dedicated “personalized_image_url” column, alongside status flags like “image_ready” or “email_sent.” That makes the sheet both log and launchpad for the next step: sequenced outreach.

For more complex branches—like retry logic, error tagging, or multi-variant testing—the n8n Documentation shows how to add conditional nodes and multiple downstream paths. At scale, this assembly line quietly turns a cold CSV of 500 leads into 500 on-brand, Nano Banana–powered visuals ready to drop into your outbound campaigns.

Unleash Your AI Marketing Co-Pilot

Cold outbound images are just the opening act. Once Nano Banana and n8n start talking to each other, you effectively get a visual content engine that can touch every part of your marketing stack, from social feeds to YouTube to ad platforms.

Start with social. Your CRM or Airtable can trigger Nano Banana to generate a new image every time you publish a blog post, close a deal, or launch a feature. n8n pushes those assets straight into a scheduler like Buffer or Hypefury, so LinkedIn, X, and Instagram get a constant stream of on-brand, auto-personalized visuals without a designer touching Figma.

Ad creatives become a data problem, not a design bottleneck. Feed your product catalog, audience segments, and value props into a JSON prompt template, then let Nano Banana output dozens of variants per audience: different headlines baked into the image, swapped backgrounds for regions, or industry-specific visual metaphors. Run that through Meta or Google Ads APIs and you are A/B testing 20–50 image concepts per week instead of 2–3 per quarter.

YouTube thumbnails might be the most obvious cheat code. Pull video metadata from the YouTube API, generate a punchy title with ChatGPT, then have Nano Banana recreate a proven thumbnail structure: host cutout, bold text, brand logo, and a high-contrast background. n8n can automatically upload the thumbnail via API, so every new video ships with a conversion-optimized visual by default.

The real play is treating this as a modular AI co-pilot. Swap “lead list” for “newsletter subscribers,” “podcast guests,” or “webinar registrants,” then adjust the prompt and destination: social posts, thumbnails, ad sets, even personalized onboarding assets.

You already have the stack: Nano Banana, n8n, a data source, and an email or social tool. Build one automated image workflow this week, ship it to a small audience, measure the click-through rate, and then decide whether your next hire should be a designer—or another workflow.

Frequently Asked Questions

What is Nano Banana Pro?

Nano Banana Pro is a powerful AI model designed for generating high-fidelity images. It excels at incorporating text and logos, making it ideal for creating personalized branding and marketing content.

How much does this automation cost per lead?

The core image generation costs around 10 cents per image via the Kia API. Costs for other services like n8n, Apify, and data enrichment tools will vary based on your usage volume and subscription tiers.

Do I need to be a developer to build this?

No. This entire workflow is built using no-code and low-code tools like n8n and Apify. It's designed for tech-savvy marketers and business owners, not just developers.

Why is n8n the right tool for this workflow?

n8n acts as the central 'brain' of the operation. It connects all the different services (lead scraping, AI prompting, image generation, cloud storage) into a single, seamless, and automated workflow.