How to Create a Comic Strip with Nano Banana — Locked Characters, Consistent Panels

Comic artists have always had a problem with AI image generators: generate the same character twice and you get two different people. Different nose, different jacket, different energy entirely. Nano Banana — Google’s Gemini Flash Image generator — tackles this head-on with subject consistency mode, which pins your characters’ faces, clothing, and expressions across multiple generated panels. The result is something that actually resembles a coherent comic strip rather than a casting call for five different actors playing the same role.

This tutorial walks through the full workflow: defining your characters, locking them in, building panel sequences, and getting clean multi-scene output that a real comic artist would actually want to use. Whether you’re making a three-panel gag strip or a longer story sequence, these techniques apply directly. Bring your prompts and let’s get into it.

What You’ll Achieve

By the end of this tutorial, you’ll know how to generate a multi-panel comic strip where the same character appears consistently across scenes — same face, same outfit, same visual identity. You’ll also have a working library of ready-to-paste prompts covering different panel types, shot compositions, and emotional beats, plus the pro tips that separate passable AI comics from ones people actually want to read.

What You Need Before Starting

Access to Nano Banana via the Gemini app (gemini.google.com), Google AI Studio (aistudio.google.com), or the Gemini API. The image generation capability runs on Gemini Flash Image underneath, so any of these access points gets you the same model. For multi-panel consistency work, AI Studio is the better choice — it gives you more control over parameters and lets you keep a running context window that helps the model remember your character definitions. A free Google account gets you started; Gemini Advanced or API access gives you higher resolution output and more generations before hitting rate limits.

Step 1 — Define Your Character in a Master Prompt

Subject consistency lives or dies at this stage. Before generating a single panel, write a detailed character reference prompt. This becomes your anchor — every subsequent panel prompt references back to it. The model needs specifics: hair color and style, eye color, skin tone, age range, and most importantly, the exact clothing your character wears. Vague descriptions produce drift. Specific descriptions hold.

Here’s a solid master character prompt to establish a protagonist:

Character reference sheet, comic book style. Young woman, mid-20s, short curly auburn hair, green eyes, light brown skin, wearing a worn olive-green bomber jacket with a red patch on the left shoulder, black jeans, white sneakers. Confident expression. Flat color illustration style, bold black outlines, warm color palette. Full body front view, white background, character design sheet.

Generate this first. Save it. This is your visual bible for everything that follows. The red patch on the shoulder is the kind of specific detail that gives the model a unique anchor point — generic clothing details get forgotten, distinctive ones stick.

Pro tip ✅

Add one visually unusual detail to your character — a distinctive accessory, an asymmetric haircut, a specific pattern on their clothing. These serve as consistency anchors that the model latches onto far more reliably than generic descriptors like “brown hair” or “casual clothes.”

Step 2 — Build the Panel Sequence Structure

Decide your comic format before generating panels. A three-panel strip (setup, complication, punchline) is the easiest starting point. A six-panel page gives you room for actual storytelling beats. Plan your shot types before you prompt — wide establishing shots, medium two-shots, close-up reaction panels. Writing these out in advance means you’re directing a comic, not just generating random images and hoping they cut together.

For a three-panel strip introducing your character arriving at a strange location, your shot list might look like: Panel 1 — wide shot, character walking toward a mysterious door; Panel 2 — medium shot, character looking surprised at something off-panel; Panel 3 — close-up reaction, comic expression of disbelief. Now translate those into prompts.

Step 3 — Generate Panels with Consistent Character References

Each panel prompt needs to call back to your character definition explicitly. Don’t assume the model remembers from the previous generation — restate the key identifiers every time. Use the phrase “same character as reference” and re-list the two or three most distinctive visual traits.

Panel 1 — establishing shot:

Comic strip panel 1 of 3. Same character: young woman, short curly auburn hair, olive bomber jacket with red shoulder patch, black jeans. Wide shot. Character walking toward a large ornate wooden door on a foggy street at night. Flat color comic style, bold black outlines, warm amber street lighting. Horizontal panel format, 16:9.

Panel 2 — medium reaction shot:

Comic strip panel 2 of 3. Same character: young woman, short curly auburn hair, olive bomber jacket with red shoulder patch. Medium shot, waist up. Character looking surprised, eyebrows raised, mouth slightly open. Door now open behind her, strange blue light spilling out. Same flat color comic style, bold black outlines. Horizontal panel format, 16:9.

Panel 3 — close-up expression:

Comic strip panel 3 of 3. Same character: young woman, short curly auburn hair, olive bomber jacket with red shoulder patch. Close-up face shot. Expression: wide-eyed disbelief with slight smile, comic exaggeration style. Blue light reflecting on face. Same flat color comic style, bold black outlines. Horizontal panel format, 16:9.

Pro tip ✅

Always specify “same flat color comic style, bold black outlines” (or whatever your style anchor is) in every single panel prompt. Visual style consistency matters just as much as character consistency — a panel that suddenly shifts to painterly realism breaks the strip completely, even if the character looks right.

Step 4 — Adding a Second Character Without Losing the First

Two-character scenes are where most AI comic workflows fall apart. The trick is to define Character B with equal specificity in a separate reference prompt first, then combine both character descriptions in the shared-panel prompt. Keep their descriptions clearly differentiated — if both characters have similar builds, give them visually opposite color palettes.

Two-character scene prompt:

Comic panel, two characters. Character A: young woman, short curly auburn hair, olive bomber jacket with red shoulder patch, black jeans. Character B: tall older man, grey beard, wire-rimmed glasses, dark blue trench coat, brown scarf. Both characters facing each other in medium shot, Character A looking suspicious, Character B looking amused. Interior setting, dimly lit library. Flat color comic style, bold black outlines, cool blue-green color palette. Horizontal panel, 16:9.

The more distinct the two characters look from each other, the better the model holds both of them. Avoid generating two characters with identical body types and similar hair unless you want the model to occasionally swap their clothes between panels.

Warning ⚠️

Nano Banana’s subject consistency works best with up to three distinct characters in a scene. Beyond that, expect degradation — clothing details start migrating between characters, and facial features blur. For ensemble casts, focus each panel on a subset of characters rather than trying to fit everyone in frame at once.

Step 5 — Text Rendering for Speech Bubbles

Nano Banana handles short text better than most AI image generators, but comic lettering inside speech bubbles is still an area where results vary. For best results, keep bubble text short (under eight words), use all-caps lettering instructions, and specify the bubble shape explicitly. Alternatively, generate clean panels without text and add lettering in a separate tool like Canva, Clip Studio, or even Google Slides — this is what most serious comic workflows do regardless of the generator.

Panel with speech bubble:

Comic panel, medium shot. Young woman, short curly auburn hair, olive bomber jacket with red shoulder patch. Character pointing dramatically, speech bubble containing ALL-CAPS text: "THIS IS DEFINITELY FINE." Oval speech bubble, black outline, white fill, positioned upper right. Flat color comic style, bold black outlines. Horizontal panel, 16:9.

Action panel with sound effect text:

Comic panel, wide shot. Young woman, short curly auburn hair, olive bomber jacket, running at full sprint through a crowded marketplace. Large bold ALL-CAPS sound effect text: "WHOOOOSH" in dynamic diagonal placement, yellow and red colors, comic impact style. Flat color illustration, bold outlines, high energy composition. Horizontal panel, 16:9.

Note 💡

Every image Nano Banana generates carries a SynthID watermark — Google’s invisible AI content identifier embedded in the pixels. You won’t see it, but it’s there. For commercial comic work, check the current Gemini API terms of service regarding watermarks and usage rights before publishing.

Step 6 — Different Visual Styles for Different Comic Genres

The style anchor in your prompts completely changes the genre feel of the output. Swapping the style descriptor while keeping the character description identical is one of the most useful creative tools in this workflow. Here are four style variants that produce distinctly different comic aesthetics with the same character:

Manga-style panel:

Manga comic panel. Young woman, short curly auburn hair, olive jacket with red shoulder patch. Close-up emotional reaction, large expressive eyes, manga-style speed lines in background, dramatic lighting. Black and white with screentone shading, sharp ink line art, classic shonen manga aesthetic. Horizontal panel, 16:9.

Retro newspaper strip style:

Vintage newspaper comic strip panel, 1950s style. Young woman character, auburn hair, olive jacket. Three-quarter view, medium shot, standing on a suburban street. Hatched shading, limited two-color palette (black and tan), worn print texture, classic newspaper funny page aesthetic. Horizontal panel format.

Gritty noir graphic novel:

Noir graphic novel panel. Young woman, short curly auburn hair, olive bomber jacket with red shoulder patch. Standing in rain under a streetlamp, high contrast shadows, rain streaks visible. Black, white, and single red accent color only. Frank Miller influence, heavy ink shadows, cinematic composition. Horizontal panel, 16:9.

Bright all-ages adventure comic:

All-ages adventure comic panel, bright Saturday morning cartoon style. Young woman, short curly auburn hair, olive jacket with red shoulder patch. Action pose, jumping forward with determined expression. Flat bright colors, thick outlines, clean simple backgrounds in pastel blue sky. Cheerful energetic composition. Horizontal panel, 16:9.

Pro tip ✅

Pick your style anchor on panel one and never deviate from that exact phrasing. “Flat color comic style, bold black outlines” and “flat illustration with thick outlines” will produce slightly different results even though they sound identical. Copy-paste your style descriptor verbatim across every panel prompt — don’t retype it from memory.

Step 7 — Assembling the Final Strip

Nano Banana generates individual panels, not assembled strips. Once you have your panel images, bring them into any layout tool to assemble the final comic. Google Slides works fine for quick results. Canva has comic strip templates that handle gutters and panel borders automatically. For anything more serious, Clip Studio Paint or even Adobe Express handles the assembly cleanly.

Export each panel at the highest available resolution — use “4K resolution” or “high detail” in your prompts to push the output quality up before you download. Panels generated at higher detail hold up better when you print or display them at larger sizes.

Pro tip ✅

Generate a few extra panels beyond what you need. AI consistency isn’t perfect — sometimes panel three has a character whose jacket suddenly turned grey. Having backup generations for each panel means you’re selecting the best of three attempts rather than living with whatever the first generation gave you.

Why This Workflow Actually Holds Together

The reason this approach works — and why it’s meaningfully different from just running generic image prompts — is the combination of explicit character anchoring, style consistency, and shooting to a pre-planned panel sequence. Each of those three elements does a specific job. Skip one and the others can’t compensate. Write a detailed character reference and restate the key identifiers every panel, lock your visual style phrasing and paste it identically each time, and plan your shot sequence before you generate anything.

Independent comic creators using Nano Banana for short-form strips report the biggest time savings come from concept-to-rough-layout speed — not having to draw character studies and environmental sketches from scratch lets artists spend their actual time on story and composition decisions. The AI handles the visual labor of “make this character look like that character in this setting”; the human handles the craft of whether the strip is actually funny, moving, or worth reading. That division of labor is exactly where AI image generation pays off — and Nano Banana’s consistency tools make it a credible option for this workflow in a way that earlier generators simply weren’t.

promptyze

See Full Bio