Hi, how can we help you today?

Edit with Prompts


Introduction

"Edit with Prompts" allows users to modify images using simple text instructions.

It uses advanced AI models like GPT-Image (also known as GPT-4o images) and Gemini 2.0 to interpret natural language prompts and apply the described changes directly to one or more images.

The feature is designed to simplify and accelerate the editing process, avoiding the need for prompting traditional diffusion models like SDXL or Flux. Being fully integrated into the Scenario AI toolsuite and API, "Edit with Prompts" works seamlessly alongside other tools to streamline image editing and give users more control.

A series of character all edited with the same prompt: “Remove the sword, and change character pose to a dab dance with arm extended and head lowered“


Capabilities

To use Edit with Prompts, simply describe the changes you want to make with clear text instructions, and the AI will apply those edits directly to your image. The tool supports a wide range of powerful transformations, including:

  1. Object Addition & Removal: Add new elements to the scene or remove existing ones.

  2. Object Modification: Change properties like color, texture, size, or style of objects already in the image.

  3. Background Manipulation: Modify the background, such as adjusting the time of day, weather, setting, or mood.

  4. Style Transfer: Apply a different artistic style—for example, converting a photo into an oil painting, pencil sketch, or cartoon style.

  5. Scene Composition: Combine elements such as placing a character into a vehicle, merging objects or creatures, or designing environments like rooms, landscapes, or interiors.

  6. And much more!

The quality and relevance of the generated output will depend on several key factors, such as the reference image(s) you provide, the clarity and specificity of your instructions, the AI model selected (e.g., GPT-Image or Gemini) and any additional settings.

Overall, Edit with Prompts is built to accelerate creative workflows, helping you generate high-quality visuals more quickly, with greater flexibility and control.

Instructions: Create an isometric version of the beach house in the photo, following the style of the 5 provided isometric images (3D cartoon style). Keep a plain, neutral background


Access "Edit with Prompts" 

You can launch the Edit with Prompts tool in several ways within Scenario:

  • From an individual image: Open the image you want to edit, click the three-dot icon in the top-right corner, and select "Edit with Prompts". This will open the tool with the image already loaded into the interface.

  • From any gallery view: Whether browsing generated images or exploring your library, hover over any image thumbnail and click the three-dot icon in the top-right corner to find the "Edit with Prompts" option.

  • Via the main navigation panel: Go to Images → Edit with Prompts to open the tool from scratch. No image will be pre-loaded. You can also find it under the "All Tools" page.

  • Direct link: Access the tool directly via this URL: https://app.scenario.com/edit-with-prompts


Interface Overview

The Edit with Prompts interface is split into two main sections. Below is an overview of the key components in each area.

  • A (left-hand panel) is for input and configuration

  • B (right-hand panel) is for reviewing, managing, and reusing outputs.


A. Left-Hand Panel: Inference and Control Configuration

This side of the interface contains all the tools needed to set up your image editing request.

  1. AI Model Selection

    Choose between the available AI models: GPT-Image or Gemini 2.0 Flash Experimental. Gemini is great for precise edits, like adjusting small details or swapping specific elements while preserving most of the original image. GPT-Image is better for more creative or broad edits, such as changing character proportions or restyling an entire scene. Each model behaves differently, so it’s helpful to experiment and compare.

  2. Reference Image Upload

    Upload the image you want to edit using several supported methods:

    • Local Upload: Select an image file from your computer.

    • Drag and Drop: Drag a file directly into the upload area (you may drag and drop from the right panel directly into the left one).

    • Paste from Clipboard: Paste an image you’ve copied.

    • Scenario Workspace/Library: Browse and select images stored in your Scenario workspace or team library. You can filter by tags or collections for faster access.

    • Once an image is uploaded, you may remove its background, crop it, or even use the sketch tool to draw directly on the image and guide the AI for more controlled edits

  3. Instructions (Prompt Box)

    This is where you type your prompt: the natural language instructions describing the changes you want to apply to the image. The more specific and clear your prompt, the better the results will be.

  4. Prompt Assistance Tools

    Scenario includes built-in helpers to make writing effective prompts easier:

    • Generate Suggestions: Based on your uploaded reference image, Scenario suggests several prompt ideas. Click any suggestion to insert it directly into the prompt box and edit as needed.

    • Rewrite Your Prompt: Automatically rephrase your current prompt for better clarity or structure - even if it’s just a few words.

    • Translate Prompt: Write your prompt in any language, and Scenario will translate it to English automatically for optimal results

  5. Additional Images

    You can upload up to five extra images to guide the model. These images help set style, composition, or content references. Be sure to describe them in your prompt (e.g., “match the style of the provided forest landscape”) instead of referring vaguely (e.g., “use image 2”).

  6. Settings & Parameters

    Depending on the selected model, you’ll have access to adjustable settings that influence the output:

    • Image Count: Set how many variations to generate.

    • Aspect Ratio: Choose the output’s dimensions or proportions.

    • Quality (GPT-Image only): Control the level of detail and visual fidelity in the generated image.


B. Right-Hand Panel: Output Review & Iteration 

The right-hand side of the Edit with Prompts interface displays all your generated outputs. It’s designed for easy review, comparison, and reuse in the upcoming steps of your workflow.

  1. Layout Options

    Choose how you want to browse results: grid view, row view, or other layout styles. This makes it easier to scan multiple outputs and compare them side by side.

  2. “Compare” Slider

    When viewing a generated image, activate the side-by-side comparison slider (keyboard shortcut: C) to directly compare the edited version with the original reference image.

  3. Reuse Generated Images (Drag & Drop)

    You can drag any generated image from the output panel back into the reference or additional image slots. This makes it easy to continue editing or build on previous results without needing to upload again.

  4. Reuse Prompts & Settings

    Click the three-dot menu on any output image to quickly reload its original prompt, reference image, and generation parameters: ideal iterating based on a previous result.

The interface is built to support a fluid, iterative workflow. It allows users to adjust inputs, generate new results, review outputs, and refine their prompts or settings in a continuous editing loop.


Step-by-Step Guide

Step 1: Select the AI model that best fits your needs.

  • Gemini 2.0 Flash Experimental: Ideal for precise, focused edits. It makes small, localized changes while keeping the rest of the image mostly intact. Great for tasks like changing colors, swapping accessories, or removing elements without affecting the background or overall style.

  • GPT-Image (4o images): Best for more creative or wide-ranging edits. This model can restyle characters, change lighting or backgrounds, and make bigger transformations across the whole image. Use it when you want a more dramatic or artistic result.

Tip: If you’re not sure which model to use: try your prompt with one, then switch to the other with a single click to quickly compare results and see which works best for your needs.

Step 2: Upload the image(s)

You can use a single image (Reference Image), or optionally add up to five Additional Images using the panel below to provide extra visual guidance.

Step 3: Write a clear instructions

In the instruction box, type a clear, specific description of the changes you want to make to the image(s). Here are some best practices:

  • Be Specific: Avoid vague prompts like “make it look better.” Instead, say something like “change the sky to a stormy gray with lightning bolts.”

  • Keep It Focused: Especially when starting out, stick to one or two main changes per prompt. Simpler prompts often produce more predictable results.

  • Use Descriptive Language: Include details about style, texture, mood, or features (e.g., “a glowing blue aura,” or “an aged wooden table with a dark finish.”)

  • Refer to Elements Clearly: Identify specific parts of the image, such as “the character on the left” or “the red apple in the bowl.”

  • If Using Additional Images: Mention how each one should influence the result - for example, “apply the style of the uploaded Van Gogh painting to the background,” rather than vague references like “use image 1’s style.”

  • Iterate as Needed: If the results aren’t quite right, tweak your prompt: add details, rephrase, or simplify… and try again.


Integrating into Scenario Workflows

“Edit with Prompts” is fully integrated throughout the Scenario platform and works seamlessly with other AI tools. You can launch it from various points in the app, and it becomes especially powerful when used as part of a broader, multi-step creative workflow.

You can use it before or after other actions, whether you’re editing an initial image, enhancing a result, or refining outputs from other tools.

Flexible Handoff Between Tools

Any image, whether generated from a base model, a custom-trained model, or edited with other tools, can be instantly sent to “Edit with Prompts” for further modification using natural language. After editing, you can continue refining the result using complementary tools such as:

  1. Expand for outpainting and extending the image

  2. Enhance for upscaling and improving the quality/sharpness 

  3. Retouch for targeted inpainting on the Scenario canvas

Or you can recover style consistency by using your custom-trained model with image-to-image. This flexibility helps you maintain creative control even if edits by GPT Image or Gemini change the image more than expected.

Composition and Assembly Workflows

“Edit with Prompts” is also useful for combining elements created separately in Scenario. For example:

  1. Combine a character and outfit created on separate canvases into one cohesive image

  2. Merge a vehicle with its “driver“

  3. Place objects, creatures, or accessories into a shared scene

This makes it especially valuable for building scenes and assembling complex compositions.

Variant Generation and Model Training

You can use “Edit with Prompts” to generate multiple variations and support model training workflows:

  • Create different versions of a character with varied poses, outfits, backgrounds, or styles

  • Use these outputs directly to train a custom model with one click using the "Train a model" button

If you have a dataset that needs a few more images in the training set, you can use “Edit with Prompts” to generate additional material for a better training.

Organize 

And as with all the Scenario tools, you can always use collections and tags to further organize the outputs, all in a collaborative setups.


Illustrative Examples

Here are a few simple examples of how you can use "Edit with Prompts" with GPT-Image or Gemini 2.0 Flash Experimental. The outcome depends on the base image, the clarity of your prompt, the selected model, and any generation settings.

These examples are just a starting point… there are hundreds of ways to use this tool. Try your own images, write your own prompts, and experiment freely. The tool is built for creative exploration, not rigid steps.


1. Change the outfit of a character

Instructions: “Change the outfit to the medieval outfit shown in the reference.”


2. Change a pose 

Instructions: “Remove the sword, and change character pose to a dab dance with arm extended and head lowered”


3. Add a background 

Instructions: “Place the character in a dungeon similar to the provided background, with a cartoon style”


4. Combine Elements

Instructions: “Create an image of this character wearing the exact armor shown and holding the same weapon provided in the reference images”


5. Explore Facial expressions

Instructions: “Generate a 6-expression sheet based on the input character headshot and the expression sheet provided (sketches). Maintain character likeness and style. Add one single word under each expression to describe it, using a hand-written style font “


6. Explore hairstyles

Instructions: “Create a single image showcasing six different hairstyle variations for the character, inspired by the provided hairstyle sheet. Keep the character's facial features and style consistent while varying the head poses. Use a white background.”


7. Render a sketch

Instructions: “Reposition the armored character (comic style) from the provided image to match the pose shown in the line art sketch exactly (dynamic jumping pose). Ensure the comic style remains consistent. Align all limbs and body positioning with the sketch, including arms extended on each side as if flying, head leaning forward, and one knee bent”


Conclusion

"Edit with Prompts" brings the power of advanced AI models like GPT-Image and Gemini into a fast, intuitive editing experience, driven entirely by natural language. 

Its deep integration with Scenario’s tool suite makes it more flexible and efficient than standalone alternatives, helping you streamline complex creative workflows.

Was this helpful?