Magic Canvas

Create and iterate on images and videos in a persistent project workspace. Generate multiple variations, compare outputs side by side, and explore creative directions; all within a single organized canvas.

What you can do:

Generate up to 15 image variations from a single prompt
Create up to 2 video variations per prompt
Organize work in projects with multiple canvases
Use templates for quick generation or customize settings
Compare, regenerate, and refine outputs in one workspace
Combine images with "Mash image" feature

Use Magic Canvas for creative exploration that requires iteration, comparison, and experimentation.

Quick Start

Create a project to organize your work
Choose a canvas or create multiple canvases for different ideas
Enter a text prompt describing your image or video
Generate variations (up to 15 for images, 2 for videos)
Compare outputs on the canvas and select your favorites
Iterate by regenerating, creating more variations, or mashing images

tip

Start with a single canvas and simple prompt. You can always add more canvases or refine your prompts as you explore.

Use Cases

Magic Canvas excels at creative workflows that benefit from iteration, comparison, and exploration.

Ideal for:

Creative brainstorming and concept exploration: Generate multiple variations to explore different visual directions before committing to a final concept
Marketing campaign variations: Create and compare different versions of marketing visuals, social media content, or ad creatives
Storyboarding and video concepts: Develop video ideas by generating variations and comparing motion, pacing, and composition
Product visualization iterations: Explore different product presentations, backgrounds, and styling options
Design system exploration: Test multiple design directions, color schemes, or compositional approaches in parallel
Content series creation: Generate related images or videos that share a common theme or style

Best results when:

You need to compare multiple creative options
Iteration and refinement are part of your workflow
You want to explore variations without losing previous work
You're working on related content that benefits from shared context

Projects in Magic Canvas

Work in Magic Canvas is organized into projects. Each project acts as a container for:

Text prompts and generated outputs
Image and video variations
Reference images
Multiple canvases for different ideas

Projects enable iterative workflows where you can continue generating and refining content within the same workspace.

Multiple Canvases

Each project can contain multiple canvases, independent working surfaces for exploring different ideas without mixing outputs.

Use multiple canvases when:

Exploring different creative directions in parallel
Separating image and video experiments
Iterating on variations without cluttering a single canvas

Using multiple canvases is useful when:

Exploring different creative directions in parallel
Separating image and video experiments
Iterating on variations without cluttering a single canvas

All canvases within a project share the same project context but maintain separate generated outputs.

Templates in Magic Canvas

Magic Canvas supports the use of templates for both image and video generation.

Templates provide predefined configurations that may include:

Model selection
Prompt structure
Reference inputs
Other generation parameters

Templates allow users to start generating images or videos quickly without configuring every option manually.

Using Templates

When a template is selected, its predefined settings are applied automatically. Users can generate outputs immediately or modify the prompt and settings before generation.

Templates act as starting points rather than fixed workflows. Any changes made override the template defaults for that generation.

Image Generation

Magic Canvas enables image generation from text prompts with support for multiple variations and reference inputs.

Prompt-Based Image Generation

Enter a text prompt describing the image you want to generate. The prompt may specify the subject, environment, style, mood, or composition.

Example prompts:

A modern minimalist living room with large windows and natural light

Futuristic robot character, detailed mechanical design, studio lighting

Abstract watercolor landscape with mountains and sunset colors

Each generation is unique, giving you creative variations to explore and compare.

Image Variations

From a single prompt, you can generate up to 15 image variations.

Variations represent different interpretations of the same prompt and are useful for exploring creative alternatives without rewriting the prompt. Each variation is treated as an individual generated image.

tip

Generate multiple variations at once to compare creative options. You can always regenerate individual images or create "More like this" variations from your favorites.

Upload Image

In addition to generated outputs, users can upload their own images directly into the canvas. Uploaded images behave differently from images generated within Magic Canvas. An uploaded image can be used in two primary ways:

As a reference for generation: Uploaded images can act as visual references for image generation. When used this way, the image helps guide aspects such as style, color treatment, or visual direction for newly generated outputs. Reference images influence generation but do not become part of the generated output themselves.
As an editable canvas asset: Uploaded images can also be treated as independent canvas assets. In this mode, users can apply image transformations directly to the uploaded image without generating new images. Edits applied to uploaded images affect only that asset and are distinct from images generated through prompt-based workflows.

Model Selection for Images

Users can select a specific model for image generation or choose Auto to let the system select an appropriate model.

Available options include:

Nano Banana
Nano Banana Pro
Seedream 4.5
Auto

tip

Selecting Auto is useful when users are unsure which model best fits their prompt or desired output.

Example: Image Generation with Variations

The following example illustrates how Magic Canvas generates multiple image variations from a single prompt within a project.

Prompt used:

Stylized game character concept art for an indie game, simplified shapes, strong readable silhouette, exaggerated proportions, expressive face, limited color palette, soft painterly shading, whimsical but confident pose, clean concept art presentation

Image Credit Consumption

Credits are consumed per generated image, including each variation.

Each generated image consumes 3 credits
When using Nano Banana Pro, each generated image consumes 11 credits

Generating multiple variations increases credit usage proportionally.

Video Generation

Magic Canvas also supports video generation using text prompts, enabling motion-based content creation within the same project.

Video generation follows the same project-based workflow as image generation.

Prompt-Based Video Generation

Users generate videos in Magic Canvas by entering a text prompt that describes the desired motion, subject behavior, environment, and visual style.

Video prompts typically focus on:

How the subject moves or changes over time
Camera behavior or perspective
Scene progression and atmosphere

Writing Effective Video Prompts

Video prompts benefit from being more explicit than image prompts. Describing motion, transitions, and timing helps guide generation more reliably than static descriptions alone.

For example:

Instead of describing only how a scene looks, describe what happens
Include cues about movement, direction, or camera behavior when relevant

Examples:

Camera slowly panning across a serene forest landscape at sunrise

Product rotating 360 degrees on a clean white background

Time-lapse of clouds moving across a city skyline

Magic Canvas uses the same prompt interpretation logic as the Video Generator tool. For detailed guidance and examples on writing effective video prompts, refer to the Video Generator documentation, which covers:

Prompt structure and best practices
Common prompt patterns for motion-based generation
Limitations and failure cases specific to video generation

Each video generation is unique, offering different creative interpretations of motion and composition.

Video Variations

From a single prompt, users can generate up to 2 video variations.

Video variations allow users to compare different interpretations of motion, pacing, or composition derived from the same prompt.

Start Image

Users can optionally provide a start image to define the opening frame of the generated video.

When a start image is used, video generation begins from the supplied image and transitions into motion.

Video Models

Magic Canvas uses the same video generation models available in the Video Generator tool. Model capabilities, supported settings, and output constraints follow the same behavior.

Video Credit Consumption

Credit usage for video generation depends on the selected model and generation settings, including duration, resolution, and audio usage when supported.

Each generated video variation consumes credits independently.

Canvas Interactions

Generated images and videos are displayed directly on the Magic Canvas. The canvas allows users to preview, manage, and iterate on generated outputs without leaving the project context.

Canvas interactions differ slightly for images and videos.

Image Actions on the Canvas

After generating images or image variations, all outputs appear on the canvas. Selecting an image reveals a set of actions for managing or extending that image.

Single Image Actions

When a single image is selected, the following actions are available:

Regenerate: Re-generates the selected image using the same prompt and configuration. The regenerated image is treated as a new output and consumes credits.
More like this: Generates three additional variations based on the selected image, with minor adjustments. Each generated variation is treated as a separate image and consumes credits.
Expand: Opens the image in full resolution. Images on the canvas are displayed in a reduced size for browsing; expanding allows detailed inspection.
Download: Downloads the selected image.
Delete: Removes the selected image from the project.

Multiple Image Selection

Users can select multiple images on the canvas:

macOS: Cmd + click
Windows: Ctrl + click

When multiple images are selected, the following actions become available:

Download all: Downloads all selected images.
Mash image: Creates a new image by combining elements from the selected images. The resulting mashed image is treated as a newly generated output and consumes credits.
Delete all: Removes all selected images from the project.

tip

Mash image combines visual elements from selected images. Experiment with different combinations to discover unexpected creative directions.

Video Actions on the Canvas

Generated videos also appear directly on the canvas. Selecting a video reveals actions specific to video outputs.

Video Actions

When a video is selected, the following options are available:

Expand: Opens the video in an expanded view for playback.
Download: Downloads the selected video.
Video info: Displays metadata for the video, including:
- Model used
- Prompt
- Duration
- Aspect ratio
Delete: Removes the selected video from the project.

Quick Reference

Specification	Details
Image variations	Up to 15 per prompt
Video variations	Up to 2 per prompt
Credits per image	3 (standard), 11 (Nano Banana Pro)
Credits per video	Varies by model, duration, resolution
Image models	Nano Banana, Nano Banana Pro, Seedream 4.5, Auto
Project organization	Multiple canvases per project
Single image actions	Regenerate, More like this, Expand, Download, Delete
Multi-select actions	Download all, Mash image, Delete all
Video actions	Expand, Download, Video info, Delete
Reference images	Supported for image generation
Start image	Supported for video generation

When to Use Magic Canvas

Magic Canvas is designed for creative workflows that prioritize exploration, iteration, and comparison.

Best for:

Multi-variation workflows: When you need to generate and compare multiple options before selecting the best one
Iterative refinement: When your creative process involves progressive refinement across multiple generations
Project-based work: When related images or videos benefit from shared context and organization
Creative exploration: When discovering the right direction is more important than a single predetermined outcome
Mixed media projects: When you need both images and videos as part of the same creative workflow

Consider other tools when:

You need a single, one-off generation without iteration
Batch processing identical transformations across many images
Applying fixed, predictable edits without variation

For those scenarios, consider the Image Generator (single generations), Batch Editor (uniform transformations), or Image Studio (predictable edits).