Maintain character consistency with Gemini and MidJourney | Alpha | PandaiTech

Maintain character consistency with Gemini and MidJourney

Workflow for generating a consistent character and placing them in different scenes using image referencing.

Learning Timeline
Key Insights

Model Performance Comparison

Gemini (specifically the model referred to as Nano Banana Pro) currently provides better photorealistic character consistency than ChatGPT or MidJourney. ChatGPT tends to produce results that look obviously AI-generated, while MidJourney's workflow is significantly more involved for similar results.

MidJourney Workflow Warning

In MidJourney, you must explicitly select 'Omni Reference' before uploading the image. Failing to do so (or just using a standard image prompt) may cause the style reference to override the facial features, resulting in a loss of character identity.
Prompts

Scene Change Prompt Template

Target: Gemini / MidJourney
[Character Description] in a coffee shop
Step by Step

Maintaining Character Consistency in Gemini

  1. Generate or locate the initial character image within the Gemini interface.
  2. Download the image to your local device to serve as the character reference.
  3. Click the image upload button (typically a 'plus' or 'image' icon) in the Gemini input bar.
  4. Select and upload the previously downloaded character reference image.
  5. Type a new text prompt describing the character in a new location (e.g., 'in a coffee shop' or 'in Times Square').
  6. Click the submit button to generate the new scene while retaining character consistency.

Configuring MidJourney for Character Consistency

  1. Save the source character image to your local device.
  2. Open the MidJourney interface.
  3. Locate and select the 'Omni Reference' option in the settings or parameters menu.
  4. Click the upload area to select the source character image.
  5. Wait for the image to attach as the reference anchor.
  6. Type the text prompt for the new scene into the prompt field.
  7. Click 'Send' or press Enter to generate the image.

More from Generate & Edit Professional AI Images

View All