Peel Documentation

Welcome to Peel, the AI-powered batch image editor that transforms how creative teams work with images. Whether you're processing product photography, creating marketing assets, or experimenting with artistic styles, Peel makes it fast and effortless. No API keys to configure, no complex setup required—just open the app and start creating immediately.

Your First Batch

Processing your first batch of images takes less than a minute. Follow these steps to experience the power of AI-driven batch editing:

1. Open Peel at banana.peel.diy
2. Drag and drop images onto the upload area, or paste directly from your clipboard
3. Type an instruction in the chat (e.g., "Remove the background") or tap a preset button for common operations
4. Hit Send and watch Peel process your images concurrently with real-time progress tracking
5. Download results individually or grab them all as a ZIP file

That's the entire workflow. No credentials to configure, no billing to set up first, no tutorials to sit through. Peel is designed to get out of your way and let you focus on the creative work that matters.

The Three Panels

Peel's interface is organized into three main areas that work together to give you complete control over your image editing workflow. On mobile devices, these appear as tabs at the bottom of the screen for easy navigation.

• Input Panel: Upload images via drag-and-drop, file browser, or clipboard paste. Add text prompts for image generation. Toggle between Batch mode (same edit on multiple images) and Single Job mode (combine images into one request).
• Tasks Panel: Your command center with a chat interface and preset buttons. Monitor progress with a real-time progress bar, elapsed time counter, and token usage statistics as jobs run. View your instruction history and reuse previous prompts.
• Results Panel: View completed images in a gallery layout with side-by-side comparison to originals. Download individual results, retry failed items, or redo completed items with different instructions. Export everything as a ZIP when you're done.

Supported Formats

Peel accepts JPG, PNG, and WebP images as input. Output images are delivered in the same format as your input, with transparent backgrounds exported as PNG. For best results, use images with dimensions under 2048px—larger images are automatically resized to optimize processing speed and quality.

Nano Banana is our friendly name for the AI pipeline that powers Peel. Under the hood, we use Google's Gemini 3 Pro Image model via Vercel AI Gateway—one of the most capable multimodal AI systems available today. We've wrapped this powerful technology in an interface designed for speed and simplicity, so you get professional-grade AI image editing without the complexity of working directly with AI APIs.

What Makes It Special

Gemini's multimodal architecture sets it apart from traditional image editing tools and even other AI models. Here's what you can expect:

• Multi-modal understanding: Gemini processes both images and text together in a unified way, enabling natural language editing that understands the semantic content of your images—not just pixels.
• Context awareness: The model understands what's in your image and can make intelligent, targeted edits. Ask it to "remove the person in the background" and it knows exactly what you mean.
• Batch-native architecture: Process dozens of images concurrently with the same instruction. Peel's infrastructure is built from the ground up for batch operations.
• Resolution flexibility: Output at 1K, 2K, or 4K resolution depending on your needs. Higher resolutions cost more tokens but deliver sharper results for print and large-format displays.
• Consistent quality: Unlike models that produce unpredictable results, Gemini maintains high consistency across batch operations, ensuring your processed images have a cohesive look.

Model Capabilities

Gemini can generate images from scratch, edit existing images, remove backgrounds, upscale resolution, transform artistic styles, and much more. You interact with these capabilities through natural language instructions—just describe what you want in plain English. The model handles the technical complexity behind the scenes.

// Example: Simple generation
"A golden retriever wearing sunglasses at the beach"

// Example: Image editing
[Upload photo] + "Remove the person in the background"

// Example: Style transfer
[Upload photo] + "Make this look like a watercolor painting"

// Example: Batch operation
[Upload 20 product photos] + "Remove BG" preset

Understanding Limitations

While Nano Banana is remarkably capable, it's important to understand its limitations. AI image generation can sometimes produce unexpected results, especially with complex instructions or unusual compositions. For critical work, always review outputs carefully. The model may struggle with very specific brand elements, exact text rendering, or highly technical subjects. When in doubt, use reference images and clear, specific instructions to guide the AI toward your desired outcome.

Peel supports two fundamental input types: existing images you want to transform, and text prompts for generating new images from scratch. Both input types flow through the same powerful AI pipeline, and you can freely mix them in the same batch. This flexibility means you can process a product catalog while simultaneously generating new hero images—all in one session.

// Basic prompt - vague, generic results
"A coffee cup on a wooden table"

// Better prompt - adds atmosphere and style
"A ceramic coffee cup with latte art, on a rustic wooden
table, morning sunlight streaming through a window,
cozy café atmosphere, shallow depth of field"

// Professional prompt - precise commercial quality
"Product photography of a minimalist ceramic coffee cup,
white background, soft studio lighting, 45-degree angle,
high-end commercial style, sharp focus"

Control the size, quality, and dimensions of your generated images with Peel's flexible output settings. Whether you need quick drafts for iteration or print-ready 4K images, these controls give you precise control over your output—all accessible from the chat footer.

Settings Location

All output settings appear in the chat footer, to the left of the message input:

┌─────────────────────────────────────────────────────┐
│  [SD ▾] [Auto ▾]  │  Type your instructions...  │ ▶ │
└─────────────────────────────────────────────────────┘
     ↑        ↑
  Quality   Ratio

Settings persist across generations until you change them. When using Custom Size, the ratio picker displays your dimensions (e.g., "400×600") instead of a ratio name.

Batch processing is Peel's core superpower and the feature that makes it indispensable for creative teams. Instead of processing images one at a time like traditional AI tools, Peel lets you upload dozens or even hundreds of images, give one instruction, and watch them all transform concurrently. This fundamentally changes how you think about AI-powered image editing—it becomes a production tool, not just an experiment.

How Batch Mode Works

Batch mode is the default processing mode in Peel, optimized for applying the same transformation to multiple images. Here's how to run a batch job:

1. Drag and drop multiple images into the Input Panel, or select them from your file browser
2. Make sure "Batch" mode is selected (not "Single Job")—this is the default
3. Type an instruction in the chat or tap a preset button—your instruction applies to all images identically
4. Hit Send and watch the progress bar as images process in parallel with real-time status updates
5. Download results individually by clicking each image, or click "Download All" for a ZIP archive

Smart Concurrency

Peel automatically manages concurrency to balance speed with reliability. The system adjusts how many images process simultaneously based on your batch size, using staggered start times to prevent server overload:

• Small batches (under 5 images): 3 images process at once with 500ms stagger between starts
• Medium batches (5-9 images): 2 images at once with 1000ms stagger for stability
• Large batches (10+ images): Sequential processing with 2000ms delays ensures reliable completion

This adaptive approach means small batches complete quickly while large batches run reliably without overwhelming the AI infrastructure.

Common Batch Workflows

Batch processing unlocks workflows that would be impractical with single-image tools. Here are some popular use cases:

• E-commerce: Remove backgrounds from an entire product catalog in one session
• Brand consistency: Add brand colors or apply consistent styling to marketing images
• Archive restoration: Upscale a library of legacy images to modern resolutions
• Editorial: Desaturate photos for a cohesive muted look across a photo essay
• Content variation: Generate multiple variations of hero images with the Duplicate preset
• Social media: Transform product shots into multiple artistic styles for different platforms

Monitoring Progress

While your batch runs, the Tasks Panel shows comprehensive progress information: a visual progress bar, elapsed time, completed/pending/failed counts, and running token usage. Each work item in the Results Panel shows its individual status so you can track exactly which images are processing, completed, or need attention.

Sometimes you need the AI to see multiple images together—for compositing, style matching, or creating something that references several visual sources. That's exactly what Single Job mode is designed for. While Batch mode treats each image independently, Single Job mode sends all your uploaded images to the AI as one unified request, giving the model full context to understand relationships between images.

Single Job Mode

Toggle from "Batch" to "Single Job" in the Input Panel to switch modes. Now all your uploaded images are sent to the AI as one combined request. The model sees everything together and can understand how the images relate to each other—perfect for tasks that require comparing, combining, or referencing multiple sources.

• Compositing: Combine a product shot with a lifestyle background seamlessly
• Style matching: Show a reference image and ask "make my photo look like this"
• Multi-source creation: Reference several images to create something entirely new
• Before/after: Show the AI an example transformation and ask it to apply the same change
• Comparison-based editing: "Make the colors in image 1 match the palette of image 2"

Example Workflows

Here are some practical examples of how to use Single Job mode effectively:

// Composite example - place product in scene
[Image 1: Product on white background]
[Image 2: Lifestyle kitchen photo]
Mode: Single Job
Prompt: "Place the product from image 1 naturally on the
kitchen counter in image 2, matching the lighting"

// Style transfer example - match visual style
[Image 1: Your photo to transform]
[Image 2: Style reference with desired look]
Mode: Single Job
Prompt: "Apply the color grading, mood, and visual style
from image 2 to image 1"

// Brand consistency example
[Image 1: Product photo A]
[Image 2: Product photo B with ideal lighting]
Mode: Single Job
Prompt: "Adjust the lighting in image 1 to match image 2"

Tips for Better Composites

When combining images, results improve dramatically with clear, specific instructions. Tell the AI exactly which elements to take from which image, how they should be positioned, and what adjustments to make for seamless integration. If your first attempt doesn't match your vision, use the Redo function to refine your prompt and try again.

Presets are one-tap transformations that appear as buttons in the chat panel, giving you instant access to common operations without typing instructions. Peel ships with 6 default presets optimized for the most frequent image editing tasks, and you can create unlimited custom presets for workflows you repeat often. Presets work seamlessly with batch processing—apply them to dozens of images at once.

// Example transform style inputs
"watercolor painting with soft edges"
"Studio Ghibli anime style"
"minimalist vector illustration, flat colors"
"cyberpunk neon aesthetic, high contrast"
"vintage 1970s film photograph"
"renaissance oil painting portrait"

// Example: "Add Watermark" preset (Direct type)
Type: Direct
Prompt: "Add a subtle semi-transparent watermark reading
'SAMPLE' diagonally across the center of the image"

// Example: "Resize for Social" preset (Ask type)
Type: Ask
Ask Message: "Which platform? (Instagram, Twitter, LinkedIn)"
Prompt: "Resize and crop this image optimally for {{INPUT}}"
Validation: text

// Example: "Set Background Color" preset (Ask type)
Type: Ask
Ask Message: "Choose background color"
Prompt: "Replace the background with a solid {{INPUT}} color"
Validation: color

Reference images let you show the AI what you want instead of relying solely on text descriptions. Sometimes a picture really is worth a thousand words—especially for nuanced visual concepts like brand aesthetics, specific color treatments, or composition styles that are difficult to articulate. In Peel, reference images are attached to custom presets and sent along with every request that uses that preset.

How Reference Images Work

When you create a custom preset, you can attach up to 3 reference images. These images are sent to the AI alongside your text instruction and input images, giving the model rich visual context for understanding exactly what you want. The AI can then reference these examples when processing your images, resulting in more consistent and on-brand outputs.

• Style guides: Attach brand guidelines, mood boards, or examples of your desired aesthetic
• Color palettes: Show the exact colors and tones you want applied to outputs
• Composition examples: Demonstrate the layout, framing, or arrangement you're targeting
• Product templates: Reference images of ideal product shots for consistent e-commerce styling
• Before/after examples: Show the AI what a successful transformation looks like

Adding Reference Images to a Preset

Attaching reference images to a preset is straightforward. Once attached, they're included automatically with every use of that preset:

1. Open the preset configuration modal from your account Settings
2. Create a new preset or edit an existing one
3. Scroll down to the Reference Images section in the configuration form
4. Upload up to 3 images (each is auto-compressed to 1MB max for optimal performance)
5. Save your preset—now every use of this preset automatically includes your references

Best Practices for Reference Images

To get the most out of reference images, choose clear, high-quality examples that strongly represent your desired outcome. Multiple references work best when they're consistent with each other—conflicting visual styles can confuse the AI. For brand work, include your most definitive style examples. For technical transformations, include before/after pairs showing exactly what change you want.

The Download Modal lets you export your processed images with full control over format, filenames, and which images to include. Access it by clicking Download All in the Results header after processing images.

How to Use

1. Open the Modal — After processing images, click the Download All button in the Results panel header. The modal appears with all your completed images listed.
2. Choose a Format — At the top of the modal, select your export format:
   • WEBP — Smallest file size, great for web use
   • PNG — Lossless quality, supports transparency
   • JPG — Universal compatibility, good compression
3. Select Images — Each image has a checkbox. Click to toggle individual images on/off, or use Select All / Deselect All at the top to quickly select or clear all images.
4. Edit Filenames — Click on any filename to edit it inline. The extension is added automatically based on your chosen format.
5. Review Sizes — Each image shows an estimated file size based on your selected format. The footer displays the total estimated ZIP size.
6. Download — Click Download ZIP to convert all selected images and download them as a single ZIP file. A progress indicator shows conversion status.

Peel has a 4MB file size limit to ensure fast, reliable processing. When you upload images larger than 4MB, a modal appears with options to compress, rename, or skip those files.

How It Works

1. Automatic Detection — When you drag and drop or select files, any image over 4MB triggers the File Size Limit modal.
2. Review Oversized Files — The modal lists each oversized file with its current size and a preview thumbnail.
3. Choose an Action for each file:
   • Compress — Automatically reduce the file size while preserving as much quality as possible. The compressed size is shown as an estimate.
   • Skip — Exclude this file from your batch and continue with the rest.
4. Rename Files (Optional) — Click the filename to edit it. This display name carries through to your results and downloads, making it easy to identify images later.
5. Confirm — Click Continue to apply your choices. Compressed files are added to your batch; skipped files are ignored.

Compression Details

• Compression targets just under 4MB to fit within limits
• Quality is preserved as much as possible—the system uses iterative compression to find the best balance
• Original files on your computer are never modified

When you only need to resize or crop images without AI editing, Peel can process them instantly in your browser—completely free, with no tokens used.

How It Works

1. Upload Images — Drag and drop your images as usual.
2. Set Output Options — Use the Quality Picker (1K/2K/4K) and/or Aspect Ratio Picker to specify your desired output. You can also use Custom Size for exact pixel dimensions.
3. Leave Instructions Empty — Don't add any text instructions or select any presets. The key is: output settings only, no AI instructions.
4. Click Resize — When resize-only mode is detected, the button changes from "Make Image" to Resize. Click it to process.
5. Instant Results — Images are resized locally in your browser and appear immediately in Results. The cost shows as FREE with "Local resize" noted.

When Does Free Resize Activate?

Free resize activates when ALL of these conditions are true:

• You have image(s) uploaded (not text prompts)
• You've changed at least one output setting (quality, ratio, or custom size)
• You have NO text instructions entered
• You have NO preset selected

What Happens During Local Resize

• Images are scaled to your specified quality tier (1K = max 1024px, 2K = max 2048px, 4K = max 4096px)
• If you selected an aspect ratio, images are center-cropped to match
• If you specified custom dimensions, images are cropped and scaled to exact pixels
• Original format is preserved (JPEG stays JPEG, PNG stays PNG)
• Maximum quality (1.0) is used—no additional compression

Peel uses simple, transparent pay-as-you-go pricing. No subscriptions, no monthly commitments, no surprise fees, no unused credits that vanish at month-end. You purchase tokens when you need them, and you use tokens when you process images. That's the entire billing model—designed to be predictable and fair for teams of any size.

Token Bundles

Choose the bundle that matches your volume. Larger bundles offer better per-image rates, making them ideal for teams with ongoing image processing needs:

Tokens never expire. Use them whenever you need them, at whatever pace works for your projects. Your current balance is always visible in the header, and you can purchase additional tokens at any time without losing your existing balance. Tokens are tied to your account, so they persist across sessions and devices.

Getting Started Free

You can try Peel without purchasing tokens—we offer a few free generations for new users to experience the platform before committing. Once you've seen what Peel can do, purchase a Starter bundle to continue processing images at scale.

Tokens are the currency of Peel's AI processing. Every image you process consumes tokens based on a simple formula: the output resolution you choose determines the cost. All operations—whether you're removing backgrounds, transforming styles, upscaling, or running custom presets—use the same token rate per image. This makes cost estimation straightforward: count your images, pick your resolution, calculate your spend.

Cost by Resolution

Higher resolution output requires more AI computation and produces larger files, which is why 4K costs more than 1K/2K. Choose the resolution that matches your end use:

All operations—Remove BG, Transform, Upscale, Desaturate, Add Brand Color, and any custom presets—cost the same per image at a given resolution. There's no premium for complex operations. The only variable is your chosen output resolution.

Tracking Your Usage

Peel makes it easy to understand exactly where your tokens go with comprehensive usage tracking:

• Your token balance is always visible in the header—watch it update in real-time as you process
• Each batch displays elapsed time, progress, and running token usage as images complete
• The Account menu shows your complete job history with token usage per session
• Token purchases are confirmed instantly and reflected in your balance immediately