The $20 Image Stack.

I went looking for the answer to the same question everyone keeps sending me. What I found was a hundred guides funneling into the same place: a subscription site selling generations.

Higgsfield. Magic Hour. ImagineArt. That paid Canva plugin. They all wrap the same models I can access for free. GPT Image 2 in ChatGPT. Gemini 3.1 Flash Image (Google's internal name: Nano Banana 2) in Google AI Studio. The ones underneath, no middleman.

So I stopped reading tutorials and started building. The challenge: produce everything on a Claude Pro subscription, nothing else.

$20 a month. A brand kit. The free models underneath. That's the whole stack.

Before the walkthroughs, the gate.

Everything below assumes you have a brand kit. Palette, fonts, voice, layouts, reference images.

Brand kit (defined): a reusable document encoding your visual identity - hex codes, font stack, tone of voice rules, and reference images - that you feed to an AI tool so it generates on-brand output instead of generic output. Without it, AI gives you generic. With it, AI gives you you.

What does the $20 AI image stack actually cover?

Five content types, one subscription:

• Single branded image - Claude Cowork + brand skill, under 3 minutes, no Canva
• Branded carousel - Claude Cowork multi-slide, HTML-rendered, exported as PDF or PNG
• Viral portrait - GPT Image 2 (ChatGPT) or Gemini 3.1 Flash Image (Google AI Studio), free, no watermark
• Infographic - three paths depending on visual style (HTML, image model, NotebookLM)
• 8-second video - Veo 3.1 in Google Flow, free with watermark, commercial use permitted

1. The single branded image

Claude Cowork with a saved brand skill produces branded single images in under 3 minutes at no cost beyond the Claude Pro subscription. No Canva required.

The exact prompt I drop in:

"Use my brisk. brand skill. Single-image Instagram post on this topic: 'You don't need a paid AI subscription yet.' Format 1080×1350 PNG. Headline short with one italic accent word. Body: 4 numbered patterns, one line each. Brand colors only. Output as artifact."

Cowork builds an editable artifact. Tweak the headline if needed, download the PNG, post.

What changed in 2026: GPT Image 2 (OpenAI, launched April 21, 2026) and Gemini 3.1 Flash Image both now render typography accurately inside generated images. The reason you used to finish text-heavy images in Canva no longer exists.

When to go to Canva vs. when to stay in Claude Cowork

• Clean grids, text overlays, structured layouts → Claude Cowork
• Hand-drawn elements, custom illustrations → image model directly
• Fine kerning control or brand asset overlays → Canva

2. The branded carousel

I personally designed a full range of carousel page types in two design variations, then encoded them as a brand skill. Claude Cowork reads the skill and renders each slide as an HTML artifact.

Full step-by-step setup in the Carousels article. Paid subscribers get the exact carousel template.

How do you make viral AI portraits of yourself without paying?

GPT Image 2 (ChatGPT free tier) and Gemini 3.1 Flash Image (Google AI Studio) both accept a reference headshot, maintain facial likeness, and generate styled portraits - Polaroid effects, 3D figurines, vinyl toy figures, magazine covers - with no watermark and commercial use permitted on the free tier.

Option A: ChatGPT free tier (GPT Image 2)

• Approximately 3 generations per day
• No watermark, commercial use OK
• Best text rendering currently available (status: May 2026)
• Best when you want one polished, final output

Option B: Google AI Studio (Gemini 3.1 Flash Image / Nano Banana 2)

• Approximately 50 generations per day
• No watermark, commercial use OK
• Best when you need to iterate fast

Setup is identical in both: upload one clean headshot, reference it in the prompt, run.

Vinyl figure prompt - swap "brisk." for your brand:

"Oversized round-headed vinyl collectible figure of the person in the reference image. Large solid black dot eyes, minimal facial features, glossy plastic finish. Place the figure inside a collector's display box with a transparent front window. Box exterior in deep teal (#00C9A7) with bold orange (#E8622A) accents. Top of box reads 'brisk.' in clean sans-serif. Beneath, smaller text: 'AI for people who don't speak tech.' Inside the box beside the figure: a tiny laptop, a coffee cup, a small Substack icon. Studio product photography lighting, clean white background outside the box. 4:5 portrait aspect ratio. Maintain exact facial likeness to reference image."

Option C: The reverse-engineering trick

Find a viral image. Drop it into Claude or Gemini. Type: "Write me a prompt that would reproduce this image with a different subject."

You get the prompt back. Swap the subject. Run it. This is the part nobody publishes because there's no affiliate revenue in it.

What are the three paths for AI infographic creation?

Three distinct approaches exist, each suited to a different visual output.

Path 1: Claude renders HTML, you screenshot it

Best for: clean grids, comparison layouts, data-heavy structures.
Not suited for: hand-drawn, sketchy, or illustrated aesthetics.

Prompt structure:

"Create a self-contained HTML file with inline CSS for an Instagram infographic, 1080×1350px. Make sure the size is screenshotable. Topic: [your topic]. Brand: [your brand name]. Colors: teal [hex] accent, orange [hex] secondary, off-white [hex] background, near-black [hex] text. Font: [your font]. Layout: bold title at top, accent word in cursive, 5 cards in a grid, footer with handle. Render as artifact."

Claude renders it. Screenshot at 1080×1350. Post.

Path 2: GPT Image 2 or Gemini 3.1 Flash Image directly

Best for: illustrated or hand-drawn infographic styles.

Use the same reference-headshot workflow as the viral portraits. Free in both ChatGPT and Google AI Studio.

Path 3: NotebookLM

Best for: research-backed infographics built from your own source material.

Upload your sources, request an infographic in your brand colors, and NotebookLM generates one from your material without further prompting. Full playbook: The NotebookLM Playbook.

A note on Claude Design

Anthropic launched Claude Design on April 17, 2026, running on Opus 4.7. Without a brand skill, the output is generic. With a defined brand skill, the results are strong. Claude Design is a brand-skill amplifier - not a brand-skill replacement. Build the skill first.

Two open-source alternatives for testing without spending Pro tokens: nexu-io/open-design and OpenCoworkAI/open-codesign on GitHub.

How do you make free AI video for social media in 2026?

Google made Veo 3.1 free for all users in Google Flow in April 2026. The free tier includes approximately 50 daily credits (~12 videos per day), 720p resolution, watermarked output, no payment card required.

Prompt I used to animate a static headshot:

"Subject in the reference image stands still while the camera slowly dollies in toward the face. Soft natural light from a window on the left. Subtle hair movement. Cinematic shallow depth of field. 8 seconds, no audio."

Upload a still image (your branded post, your headshot, anything branded). Add the motion prompt. Veo returns a clip. Trim, post.

Free AI video tools compared (May 2026)

• Veo 3.1 in Google Flow - ~12 videos per day (50 credits), watermarked, commercial use permitted
• Kling free tier - 66 credits, watermarked, no commercial use
• Hailuo / MiniMax free tier - trial plus small daily bonus, watermarked, no commercial use

For commercial-safe free video, Veo 3.1 in Google Flow is currently the only option.

For animated posts that aren't camera-on-photo - text reveals, infographic motion, branded explainers - Blender (always free, open-source) and Cavalry (free since April 2026 after Canva acquisition) are the tools under active testing. Full breakdown coming in a separate article.

The full free AI image stack (May 2026)

The list nobody publishes, because there's no affiliate revenue in it.

Truly free, ongoing, commercial use permitted

• ChatGPT free tier (GPT Image 2) - ~3 generations/day, no watermark, best text rendering currently available
• Google AI Studio (Gemini 3.1 Flash Image / Nano Banana 2) - ~50 generations/day, no watermark, the volume option
• Adobe Firefly - commercially safe due to licensed training data
• Ideogram - daily free credits, typography specialist
• Krea - real-time generation, 100 daily compute units
• Cavalry - fully free since April 2026 (post-Canva acquisition), motion graphics
• Blender - always free, open-source, 3D

Free but watermarked or commercial-restricted

• Google Flow / Vids (Veo 3.1) - watermarked, commercial use permitted
• Gemini App - 20 Gemini 3.1 Flash Image generations/day, SynthID watermark applied
• Kling, Hailuo (MiniMax) - no commercial use on free tier

Trial only - do not build workflows on these

• Google AI Pro and One AI Premium 30-day trials
• Higher Google Flow tiers

Every image on my feed comes from the first list. Every one.

Subscribers get the vault: 7+ trending one-shot prompts ready to run, the reverse-engineering meta-prompt, my brand kit template, infographic HTML scaffolds, and an aspect ratio cheat sheet for every platform.