Pro Model Best Practices

Based on Google DeepMind's official Gemini Pro guide by Guillaume Vernade

Go Bananas! Pro (Gemini Pro Image) moves from "fun" image generation to functional professional asset production. It excels in text rendering, character consistency, visual synthesis, world knowledge (Google Search), and high-resolution (4K) output.

Golden Rules of Prompting

Pro is a "Thinking" model. It understands intent, physics, and composition. Stop using "tag soups" and start acting like a Creative Director.

Rule	Description
Edit, Don't Re-roll	If 80% correct, use `continue_editing` instead of regenerating
Natural Language	Write full sentences like briefing an artist, not tag soups
Materiality	Describe textures: "brushed steel", "soft velvet", "crumpled paper"
Context	Add "for whom": "for a luxury cookbook" changes everything
Identity Locking	"Keep facial features exactly the same as Image 1"

Good vs Bad Prompts

Bad: dog, park, 4k, realistic

Good: "A golden retriever playing fetch in a sunny park, captured in photorealistic detail"

Advanced Prompting Techniques

Negative Prompting

Tell the model what NOT to include using the negative_prompt parameter.

Negative	Why It Helps
`no date stamp`	Pro likes to add date stamps to corners
`no text`	Clean images without text
`not rustic`	Model tends to age things
`no watermarks`	Prevents generated watermarks

json

{
  "prompt": "Product shot of a smartphone on white background",
  "negative_prompt": "no text, no date stamp, no reflections on screen",
  "model_id": "gemini-pro-image"
}

JSON Prompting

For complex scenes, use JSON-structured prompts:

json

{
  "promptDetails": {
    "description": "Coffee shop scene",
    "styleTags": ["Lifestyle", "Golden Hour"]
  },
  "scene": {
    "background": { "setting": "cozy coffee shop", "details": "morning sunlight" },
    "subject": { "description": "Person relaxed", "pose": "holding mug" }
  },
  "technicalStyle": {
    "camera": { "shotType": "Medium Shot", "depthOfField": "Moderate" },
    "lighting": { "type": "Natural, Warm" }
  }
}

Prompt Evolution Pattern

Start simple, iterate with continue_editing:

1. generate_image({ prompt: "A portrait" })
2. continue_editing({ prompt: "make it more dramatic" })
3. continue_editing({ prompt: "add studio lighting" })
4. continue_editing({ prompt: "avant-garde fashion" })

Rule: Edit, Don't Re-roll

If 80% correct, use continue_editing instead of regenerating!

Pro Model Features Summary

Feature	Description
Resolution Tiers	1K, 2K, 4K output
Extended Aspect Ratios	13 ratios including 2:3, 3:2, 4:5, 5:4, 21:9
14 Reference Images	6 with high-fidelity for identity locking
Text & Infographics	SOTA text rendering
Viral Thumbnails	Identity + text + graphics in one pass
Storyboarding	Multi-scene sequential art
Structural Control	Sketch-to-final workflows
2D↔3D Translation	Floor plans to renders
High-Fidelity Upscaling	150x150+ to 4K
Photo Restoration	Repair tears, scratches, fading
360 Turnaround	Multiple angles from single reference
Thinking Mode	Enhanced reasoning for complex prompts
Google Search Grounding	Factual accuracy with source tracking
Person Generation Safety	Configurable modes: dont_allow, allow_adult, allow_all

Pro Model Capabilities

1. Text Rendering & Infographics

Pro has SOTA (State of the Art) text rendering capabilities.

Best Practices:

Quote exact text: "headline 'YOUR TEXT' in bold sans-serif"
Specify style: "polished editorial", "hand-drawn whiteboard"
Compress data: "summarize this as a clean infographic"

json

{
  "prompt": "Make a retro, 1950s-style infographic about the history of the American diner. Include sections for 'The Food,' 'The Jukebox,' and 'The Decor.' Ensure all text is legible.",
  "model_id": "gemini-pro-image"
}

2. Character Consistency & Viral Thumbnails

Pro supports up to 5 character reference images for "Identity Locking." (Flash supports 4.)

The Key Phrase:

"Keep the person's facial features exactly the same as Image 1."

Viral Thumbnail Template:

json

{
  "prompt": "Design a viral video thumbnail using the person from Image 1. Face Consistency: Keep the person's facial features exactly the same as Image 1, but change their expression to look excited. Action: Pose person on left, pointing right. Subject: Product on right. Graphics: Bold yellow arrow. Text: 'HEADLINE' with white outline.",
  "model_id": "gemini-pro-image",
  "aspect_ratio": "16:9"
}

3. Grounding with Google Search

Pro uses Google Search for real-time data and factual accuracy.

json

{
  "prompt": "Generate an infographic of the best times to visit U.S. National Parks in 2025 based on current travel trends.",
  "model_id": "gemini-pro-image"
}

4. Advanced Editing & Restoration

Complex edits via conversational prompting:

In-painting: Remove/add objects
Restoration: Fix old photos
Colorization: Manga/B&W photos
Style Swapping: Change artistic styles
Upscaling: Small images to 4K
360 Turnaround: Multiple angles from single reference

json

{
  "image_id": 42,
  "prompt": "Turn this scene into winter. Keep the house exactly the same, but add snow and cold lighting."
}

5. High-Fidelity Upscaling

Input images as small as 150x150 can be enhanced to 4K resolution.

json

{
  "prompt": "Upscale to 4K with enhanced detail",
  "reference_images": ["small_image.png"],
  "model_id": "gemini-pro-image",
  "resolution_tier": "4k"
}

6. Photo Restoration

Restore damaged, faded, or old photographs:

json

{
  "prompt": "Restore this old photograph to pristine condition",
  "reference_images": ["damaged_photo.jpg"],
  "model_id": "gemini-pro-image"
}

7. 360 Turnaround & Multiple Angles

Generate multiple angles from a single reference:

json

{
  "prompt": "Generate 4 images showing this product from different angles: front, 45-degree, side, and back. Maintain consistent lighting.",
  "reference_images": ["product_front.jpg"],
  "model_id": "gemini-pro-image"
}

8. Multi-Purpose Reference Images

Use references for style, branding, colors, and object placement:

Purpose	Example
Object placement	"Place this product (Image 1) in a modern office"
Style transfer	"Create in the artistic style of Image 1"
Branding	"Put this logo (Image 1) on the product"
Color palette	"Use the color scheme from Image 1"

json

{
  "prompt": "Logo from Image 1 on device from Image 2, styled like Image 3",
  "reference_images": ["logo.png", "device.png", "style.jpg"],
  "model_id": "gemini-pro-image"
}

9. Dimensional Translation (2D ↔ 3D)

Convert floor plans to 3D renders, or 2D art to 3D visualizations.

json

{
  "prompt": "Based on the uploaded 2D floor plan, generate a professional interior design presentation board. Layout: Large main image at top (living area), three smaller images below (Bedroom, Office, 3D floor plan). Style: Modern Minimalist.",
  "model_id": "gemini-pro-image",
  "aspect_ratio": "3:4"
}

10. High-Resolution & Textures

Pro supports native 1K to 4K output.

json

{
  "prompt": "Craft a breathtaking mossy forest floor. Complex lighting effects and delicate textures, pixel-perfect resolution for 4K wallpaper.",
  "model_id": "gemini-pro-image",
  "resolution_tier": "4k",
  "aspect_ratio": "16:9"
}

11. Thinking & Reasoning

Pro uses interim "thought images" (not charged) to refine composition.

json

{
  "prompt": "Solve log_{x^2+1}(x^4-1)=2 on a whiteboard. Show the steps clearly.",
  "model_id": "gemini-pro-image"
}

12. One-Shot Storyboarding

Generate sequential art with cohesive narrative flow.

json

{
  "prompt": "Create a 9-part story featuring a woman and man in an award-winning luxury luggage commercial. Emotional highs and lows, ending elegantly. Identity and attire must stay consistent. Different angles and distances. Generate one at a time. 16:9 landscape.",
  "model_id": "gemini-pro-image",
  "aspect_ratio": "16:9"
}

13. Structural Control & Layout Guidance

Use sketches, wireframes, or grids to control composition.

Sketch to Final:

json

{
  "prompt": "Create an ad for [PRODUCT] following this sketch exactly. Transform to polished design.",
  "model_id": "gemini-pro-image"
}

Pixel Art & Sprites:

json

{
  "prompt": "Sprite sheet of a character doing a backflip, 3x3 grid, frame by frame animation. Follow the reference grid structure exactly.",
  "model_id": "gemini-pro-image",
  "aspect_ratio": "square"
}

Model Comparison

Feature	Standard (Flash)	Pro
Resolution	1K	1K, 2K, 4K
Reference Images	3	14 (6 high fidelity)
Text Rendering	Basic	Advanced
Web Search Grounding	No	Yes
Thinking Mode	No	Yes
Speed	~3-5 sec	~8-15 sec

Recommended Workflow

1. Iterate with Standard (Flash) - develop concept
2. Refine prompt based on results
3. Generate final with Pro - production quality

Workflow	Tools
Text & Infographics	`generate_image` with `model_id: "gemini-pro-image"`
Character Consistency	`create_character`, `generate_with_character`
Product Marketing	`create_product_reference`, `generate_with_product`
Conversational Editing	`generate_image`, `continue_editing`

Get Help

get_help({ workflow: "viral_thumbnails" })
get_help({ workflow: "text_and_infographics" })
get_help({ workflow: "storyboarding" })
get_help({ workflow: "structural_control" })
get_help({ workflow: "dimensional_translation" })
get_help({ workflow: "upscaling" })
get_help({ workflow: "restoration" })
get_help({ workflow: "reference_mastery" })
get_help({ workflow: "prompt_iteration" })

Pro Model Best Practices ​

Golden Rules of Prompting ​

Advanced Prompting Techniques ​

Negative Prompting ​

JSON Prompting ​

Prompt Evolution Pattern ​

Pro Model Features Summary ​

Pro Model Capabilities ​

1. Text Rendering & Infographics ​

2. Character Consistency & Viral Thumbnails ​

3. Grounding with Google Search ​

4. Advanced Editing & Restoration ​

5. High-Fidelity Upscaling ​

6. Photo Restoration ​

7. 360 Turnaround & Multiple Angles ​

8. Multi-Purpose Reference Images ​

9. Dimensional Translation (2D ↔ 3D) ​

10. High-Resolution & Textures ​

11. Thinking & Reasoning ​

12. One-Shot Storyboarding ​

13. Structural Control & Layout Guidance ​

Model Comparison ​

Recommended Workflow ​

Related Tools ​

Get Help ​

Pro Model Best Practices

Golden Rules of Prompting

Advanced Prompting Techniques

Negative Prompting

JSON Prompting

Prompt Evolution Pattern

Pro Model Features Summary

Pro Model Capabilities

1. Text Rendering & Infographics

2. Character Consistency & Viral Thumbnails

3. Grounding with Google Search

4. Advanced Editing & Restoration

5. High-Fidelity Upscaling

6. Photo Restoration

7. 360 Turnaround & Multiple Angles

8. Multi-Purpose Reference Images

9. Dimensional Translation (2D ↔ 3D)

10. High-Resolution & Textures

11. Thinking & Reasoning

12. One-Shot Storyboarding

13. Structural Control & Layout Guidance

Model Comparison

Recommended Workflow

Related Tools

Get Help