Reference-based character workflow

Keep the same character across scenes and poses

Start with a structured identity lock, change one variable at a time, and inspect drift before paying for higher resolution.

Identity-lock prompt prefilledNano Banana 2 · 2K · 12 creditsFailed charged runs refund automatically

Start a character consistency edit

Upload a clear identity reference, adjust the locked traits and scene, then review the exact paid cost before generation.

Generation model

Pick the engine that matches your creative brief

Describe what you want to see

Nano Banana 22K

Paid generation · 12 credits. No free generation credits. About $0.34–$0.59 with a one-time pack.

Output controls

This run

Generation modelNano Banana 2
Image quality2K
Estimated cost

12 credits / image

About $0.34–$0.59 with a one-time credit pack. No subscription required. View packs

Aspect ratio

Output format

Image quality

Set the right expectation: character consistency is a controlled workflow, not a guarantee. Compare every result with the source and reject drift early.
Portrait reference for a character consistency workflow
Reference quality

Give the model evidence for identity and design

A face-visible portrait anchors identity. Add a full-body or side view only when proportions, clothing, or profile shape matter.

  • Use a sharp front or three-quarter image with eyes, jaw, hairline, and distinguishing traits visible.
  • Add a full-body reference when height, build, outfit construction, shoes, or accessories must stay fixed.
  • Avoid contradictory ages, hairstyles, costumes, or color grading across references.
  • Check face, age, proportions, clothing, and invented details after every paid result.

Four consistency problems this workflow addresses

Same character, new scene

Lock identity and clothing while changing only location, camera, and light.

Pose and expression

Change one action or emotion while keeping facial structure and design stable.

Story panels

Reuse the same references and lock block instead of prompting each panel from scratch.

Multiple characters

Assign each reference a name, position, and separate identity constraints to reduce trait mixing.

A lower-waste paid workflow

Solve identity drift at low resolution before adding scene complexity or paying for a final.

01

Upload one strong reference

Start with a clear, neutral, face-visible image and add only references that resolve missing evidence.

02

Lock, then change one thing

Keep the same identity block and vary only pose, expression, camera, outfit, or location per test.

03

Inspect before scaling

Test at 1K or 2K, reject drift, and use 4K or Pro only when the approved direction justifies it.

Character consistency questions

What reference image works best?

Use a sharp front or three-quarter portrait with neutral light and no obstruction. Add a full-body or side view only when it provides evidence the portrait cannot.

Does Nano Banana guarantee the exact same character?

No. A precise identity lock and stable references reduce drift, but every result still needs comparison and selection. Consistency is a workflow, not a guaranteed switch.

Should I start with Nano Banana 2 or Pro?

Start with Nano Banana 2 at 1K or 2K for lower-cost reference tests. Use Pro selectively for difficult multi-character composition, typography, or a high-stakes final.

Is character generation free?

No. The prompt and guide are free to use, but generation consumes paid credits. The editor shows the exact model and resolution cost before each run.

Lock the character before changing the scene

Use the prefilled identity structure, compare the exact cost, and generate only after the reference and change request are clear.