How to Master Consistent Characters in Nano Banana Pro for Campaigns

Brands spend millions to build visual worlds around their products, then lose that equity when AI tools can’t keep a hero character on-model from one asset to the next. As of November 2025, Google’s new Nano Banana Pro image model, running inside Google Flow, is designed to fix exactly that. With up to ten reference images, object-level editing, and tight integration with Veo video, marketers can now treat “consistent characters AI” as a practical production tool instead of a happy accident.

This evergreen guide walks you through how to master consistent characters in Nano Banana Pro for campaigns: from preparing reference images and building flows in Google Flow, to creating storyboards, ad campaigns, and product videos with locked-in characters and brand style. It assumes you have access via a Google AI Ultra subscription, which unlocks Flow and the highest-fidelity outputs for professional use.

What Nano Banana Pro is (and why it matters for brand storytelling)

Nano Banana Pro is Google’s latest high‑fidelity image generation and editing model, built on Gemini 3 Pro Image. According to Google’s November 2025 documentation and launch posts, it delivers:

• 4K-capable image generation (2K and 4K resolutions vs the older Nano Banana 1024×1024 cap).
• Improved subject identity retention, so faces, clothing, and logos stay consistent across images.
• Powerful editing tools (inpainting, outpainting, background replacement, style transfer).
• Tight integration with Google Flow and Veo 3.x for AI-powered filmmaking and storyboarding.

In Flow, Nano Banana Pro becomes a visual engine for marketing workflows: you can feed it a stack of branded reference images, generate storyboards and ad stills, then hand those frames to Veo for moving footage. This is critical if you’re building campaigns around a recurring mascot, spokesperson, or stylized product renders that must be instantly recognizable wherever they appear.

How consistent characters work in Nano Banana Pro

Under the hood, Nano Banana Pro combines two families of features that matter for marketers:

• Reference-guided generation: Using multiple reference images and Gemini’s multimodal capabilities (documented for Gemini 2.5 Flash Image and Gemini 3 Pro Image), the model learns your character’s face, proportions, clothing, and style cues, then recreates them in new prompts.
• Object- and region-level editing: Inpainting, background replacement, and localized style changes let you tweak a character’s pose, outfit, or environment without breaking identity.

Practically, this means you can use up to ten reference images:

• To teach Nano Banana Pro the character (hero, mascot, product, or set of props).
• To lock in style (lighting, lens, color grading, illustration style) across a full storyboard or ad set.
• To bridge disciplines: concept art → storyboard frames → final ad stills → Veo-driven product videos, all from the same visual DNA.

Preparing your reference images for consistent characters AI

Success with Nano Banana Pro starts before you open Google Flow. The quality, coverage, and clarity of your reference set determines how consistent your characters will be.

Step 1: Define your “hero system”

For campaign use, you’re rarely just defining one face. Instead, think in terms of a hero system:

• Primary heroes: Core characters or products that must be pixel-perfect across channels (e.g., a cartoon mascot, a flagship sneaker model).
• Secondary elements: Side characters, recurring props, branded environments (e.g., a neon‑lit bar where every scene takes place).
• Stylistic frame: Illustration or photo style, color palette, typography or UI elements if they’re part of the visual language.

List these out before curating your ten references so each asset has a clear job.

Step 2: Curate up to ten high-value reference images

As of late 2025, Google’s image and video docs emphasize that Gemini-family models perform best when you provide multiple, clear examples. Within the ten‑image limit, aim for:

• 4–5 core character references: Front, three‑quarter, and profile views; at least one neutral pose and one expressive pose; clean backgrounds.
• 2–3 style references: Images that nail lighting, color grading, and texture (e.g., “2K studio beauty shot”, “flat pastel illustration”). These may or may not include the hero.
• 2–3 context references: Typical settings or compositions (e.g., over‑the‑shoulder phone shots, kitchen counter product close‑ups, billboard mockups).

File guidelines (based on Gemini image docs and Vertex AI Imagen guidance):

• Resolution: At least 1024×1024; 2K–4K works well, but avoid blurry or heavily compressed images.
• Formats: JPEG or PNG are fully supported.
• Clean framing: Minimize motion blur, occlusions, and cluttered backgrounds when the character is the focus.

Step 3: Normalize your references

Inconsistent references lead to inconsistent campaigns. Normalize your set before uploading:

• Color and exposure: Batch‑grade images so skin tones and product colors are consistent.
• Aspect ratios: Crop to a handful of ratios you’ll actually use (1:1, 9:16, 16:9). This makes it easier to match later.
• Brand assets: Ensure logos, UI, or packaging details are sharp and legible in at least a couple of references.

Setting up Nano Banana Pro inside Google Flow

Flow is Google’s AI filmmaking and visual storytelling tool, available to Google AI Pro and Google AI Ultra subscribers (Flow launched May 20, 2025, with ongoing feature updates). To use Nano Banana Pro for campaigns:

1. Activate your subscription
   Log into your Google account with an active Google AI Ultra plan. Ultra gives you higher limits, Flow access, and watermark‑free outputs that are suitable for paid media and production work.
2. Open Flow
   Navigate to Flow via Google Labs or the Flow landing page. From there, create a new project (for a campaign, create one project per hero or major initiative).
3. Enable Nano Banana Pro
   In your project settings or media tools, select the image model as Nano Banana Pro (exposed as Gemini 3 Pro Image / Nano Banana Pro in many UIs). Confirm that high‑res (2K/4K) output is enabled.

At this point, you have a Flow project wired to Nano Banana Pro, ready to ingest reference images and generate assets.

Using up to ten reference images to lock character and style

Step 1: Upload and tag reference images in Flow

Within your Flow project:

1. Upload your curated set of up to ten reference images into the Nano Banana Pro reference panel or media library.
2. Tag each image with roles like hero-front, hero-side, style-lighting, kitchen‑context, packaging‑closeup. The more semantic structure you add, the easier it is to compose prompts around them.
3. Group references into collections or “character packs” when possible (e.g., “Spring Mascot 2026”). This lets you reuse them across scenes and campaigns within Flow.

Under the hood, this is equivalent to using the Gemini File API to upload images and referencing them repeatedly in prompts, a pattern Google’s docs explicitly recommend for reuse and consistency.

Step 2: Build prompts that bind to your references

By default, text-only prompts are more “creative” than “consistent.” To anchor on your brand system, build multimodal prompts:

• In Flow’s UI: select one or more reference images in the prompt panel, then add text instructions.
• Via API (conceptually): your contents array includes several image parts (uploaded files) followed by a detailed text prompt.

Example storyboard prompt for an ad hero:

[REF: hero-front, hero-3q, style-lighting, kitchen-context]

Create a 16:9 storyboard frame for a CTV ad.
- Keep the hero character identical to the reference: same face, hair, hoodie color and logo.
- Scene: modern kitchen at sunrise, warm backlight, shallow depth of field.
- Pose: hero stands at counter, holding our beverage can near camera.
- Expression: confident smile, relaxed.
- Maintain brand color palette from style-lighting reference.
- Output: 2K resolution, cinematic framing, minimal background clutter.

The key is to repeatedly mention:

• “Identical to the reference” for identity‑critical elements (face, logo, packaging).
• “Maintain [X] from [reference]” for style elements (palette, lighting, lens feel).
• Aspect ratio and resolution so your assets drop cleanly into media specs.

Step 3: Iterate with structured variation

Instead of writing each prompt from scratch, create a small library of reusable “prompt templates” in Flow:

• Hero close-up – face and product, minimal background.
• Hero in environment – wider shot, environment cues, supporting characters.
• Social cutdown – 9:16 vertical, room for captions and UI overlays.

Keep the character and style language fixed, and only vary scene, pose, and camera moves. This template discipline is what turns Nano Banana Pro into a “brand storytelling engine” rather than a series of one‑offs.

Object editing for on-model storyboards, ads, and product videos

Nano Banana Pro’s editing tools (similar to Imagen’s inpaint/outpaint features in Vertex AI) let you refine assets instead of regenerating whole images. For campaigns, that’s how you maintain consistency under tight deadlines.

Storyboard refinement with object-level edits

Workflow:

1. Generate rough boards
   Use broader prompts to explore camera angles and beats for your story with the hero loosely tied to references.
2. Lock the hero
   For selected frames, re‑run generation with tighter prompts and more aggressively referenced hero images (e.g., include 3–4 hero references vs 1–2).
3. Edit objects, not identity
   Use Flow’s object editing tools to adjust:
   • Arm and hand positions (e.g., lifting the product closer to camera).
   • Prop designs (updating can label to the latest SKU).
   • Background clutter (removing distracting items, adding branded elements).

Because you’re editing localized regions, the model preserves the rest of the frame—including the character’s face and the overall style—much more reliably than a full redraw.

Ad stills and key visuals

For campaign key visuals (homepage hero images, OOH, high‑impact social), combine Nano Banana Pro’s 4K output with meticulous object edits:

• Generate multiple 4K variations of your selected frame with the same reference pack.
• Use object editing to:
  • Swap outfits while keeping the same face.
  • Adjust product tilt so labels face the viewer.
  • Integrate real typography or logos as separate layers (or lightly generated, then replaced in design tools).
• Export layered assets (where Flow allows) for easy retouching in Photoshop or Figma.

From stills to product video with Veo and Flow

Google’s Veo 3.x models, available via Google AI Ultra and Flow, support image‑ and storyboard‑guided video generation. To translate your consistent character stills into motion:

1. Select master frames
   Pick 3–10 Nano Banana Pro frames that define your story arc: opening, product reveal, reaction, end card.
2. Use Flow’s story/scene builder
   Arrange frames into a sequence, add text descriptions of motion (“camera pushes in,” “hero turns to look at friend”), and link to Veo 3 / 3.1 as the video model.
3. Guide Veo with reference frames
   Configure Veo to treat your Nano Banana Pro stills as first/last frames or reference stills. Flow’s Veo docs show patterns like:
   • Text + single reference frame.
   • Text + first and last frames for more control.
   • Text + image sequence for storyboard‑driven videos.
4. Iterate locally
   If Veo drifts off-model, tighten your instructions: “keep the hero’s face and outfit identical to the reference frames; do not change hairstyle or logo.” Re‑use the same ten‑image reference set to reinforce identity in any re‑generated stills.

Campaign workflow: putting it all together

To make this tangible, here’s a typical brand storytelling pipeline using Nano Banana Pro in Flow with a Google AI Ultra subscription.

This repeatable system is where the real ROI lives: once you’ve built it for one hero or brand line, you can spin up new storylines while preserving continuity.

Best practices and limitations to keep in mind

While Nano Banana Pro and Flow are powerful, they’re not magic. A few guardrails from Google’s docs and early user reports:

• Stay within the reference bandwidth: Ten references is generous, but if they contradict each other (different hair colors, drastically different lighting), expect inconsistency. Curate ruthlessly.
• Use text to “pin” identity: Always restate identity constraints in your prompts: “same face as reference,” “do not change logo,” “keep hairstyle identical.”
• Separate composition from polish: Use Nano Banana Pro + Flow to nail composition and identity first. Then refine tiny layout and typography details in your design stack.
• Mind compute and quotas: Ultra gives higher caps, but high‑res image and video generations still consume credits. Use low‑res drafts to explore ideas, reserving 2K/4K runs for shortlisted concepts.
• Respect safety and rights: Follow Google’s usage policies. Don’t train on or mimic real people’s likeness or protected IP without the necessary rights, especially for commercial campaigns.

As of late 2025, Nano Banana Pro is evolving quickly, with enterprise options via Vertex AI and deeper integrations across Google’s creative tools. Expect incremental improvements in identity locking, reference handling, and editing fidelity throughout 2026.

Conclusion: Turning Nano Banana Pro into a repeatable brand engine

Consistent characters have always been a marketer’s dream and an AI tool’s weakness. Nano Banana Pro inside Google Flow, especially for Google AI Ultra subscribers, finally makes “consistent characters AI” realistic for day‑to‑day campaign work.

To recap:

• Curate up to ten razor‑sharp reference images that encode your hero, style, and context.
• Upload and tag those references in Flow, then build prompts that explicitly bind identity and style to them.
• Use Nano Banana Pro for both generation and object editing, refining poses, props, and environments without resetting your character.
• Extend stills into Veo‑powered storyboards and product videos, guided by the same visual DNA.

The next step is to pilot a contained campaign: choose one hero, one product line, and a single channel (e.g., paid social), and run the full Nano Banana Pro + Flow pipeline from reference to final delivery. Once you’ve battle‑tested that workflow, you’ll have a scalable template for brand storytelling across your entire marketing calendar.