Rangy User Guide

What's New in v2.9

Rangy 2.9 adds five image models and a new video model, a setup guide for API keys, and newer AI models behind the prompt tools and the Agent.

Free Plan

Rangy is free to try. No credit card is required. You get 5 image generations per day with full access to all Core tab models and features.

How to activate the Free Plan

Free vs Pro

What is Rangy?

Rangy is a standalone desktop application for Mac and Windows that lets you generate high-quality AI images and videos, and interact with design platforms like Adobe Photoshop and Affinity. It provides access to 19 AI models for image generation, 16 video generation models, image restoration, upscaling, AI Prompt Studio, prompt extraction, video editing, and more — all from one app on your computer.

Why Rangy?

Rangy was built for two main reasons:

1. Freedom from platform limitations. The developer previously built an Adobe Photoshop UXP plugin (Eti Image Generator) used by thousands of users. As a standalone app, Rangy has no restrictions — it can include advanced features, automation, AI agents, and more.
2. Support for Affinity users. Many users requested a tool for Affinity. Since Affinity does not support plugins that communicate with external AI services, a standalone app was the only solution. Rangy works alongside Affinity (and Photoshop) seamlessly.

Privacy

Installation

Download

Download Rangy from rangy.ai/download. Choose Download for Mac or Download for Windows.

Mac Installation

Windows Installation

First Launch

Updates

Rangy checks for updates when it starts and every few hours. When a new version is available, a window shows what’s new and the download size. Click Update to download it (with speed and time left); Rangy then installs it and reopens on the new version. If generations are still running, Rangy waits for them to finish before restarting, so you never lose a paid result.

Click Later to hide an update for 3 days — an Update available button stays in the header so it’s one click away. To check yourself, use Rangy > Check for Updates… on Mac or Help > Check for Updates… on Windows.

App Layout

Rangy is divided into three main columns:

Left Column: History

The history panel shows all images and videos you have previously generated. It has two main tabs:

Photos & Videos Tabs

Switch between Photos and Videos using the media tabs at the top of the history panel. Each has its own separate favorites, trash, and folder systems. Switching between Core/Pro tabs automatically shows Photos; switching to the Video tab shows Videos.

Project Folders

Organize your work into project folders. The folder bar appears below the media tabs:

• Create folders — Click the "New" button to create a new folder
• Rename/Delete — Right-click a folder for options
• Move items — Move images or videos between folders
• Generate into folders — Select a folder before generating to save results directly there
• Separate folder systems for Photos and Videos

History Features

• Favorites — Click the star icon on any image/video to mark it as a favorite. Favorites are saved in a separate folder on disk.
• Trash — Deleted items are moved to trash and kept for 30 days. Separate trash for photos and videos.
• Sorting — Items are sorted by date (newest first by default).
• Open folder — Click the folder icon to open the storage folder on your computer.
• Selection mode — Click "Select" to enter drag-selection mode. Select multiple items and delete in bulk.
• Single click — Copies the image to your clipboard for easy pasting.
• Double click — Opens the image in full-screen viewer, or video in the video player.
• Drag and drop — Drag images from history into the input image area, video resource slots, or into other apps.
• Scroll memory — Your scroll position is remembered when switching between Photos/Videos tabs.

Right-Click Context Menu

Right-click any image in history to access quick actions:

Center Column: AI Generation

This is the main workspace, containing:

• Prompt input — Write your text prompt describing what you want to generate or edit.
• AI Modify Prompt — Click the AI icon at the top-right of the prompt input to modify your existing prompt with AI. A secondary input field appears where you describe what you want to change (e.g., "make the background darker"). Claude Sonnet 5.5 or Kimi K3 (your choice) analyzes both your current prompt and your modification request, then rewrites the prompt accordingly.
• Prompt library — Click the icon at the top-right of the prompt field to access your previous prompts. You can copy, reload, or favorite any prompt.
• Reference images — Click the plus icon to add reference images for models that support them.
• AI model selection — Choose one or more AI models to generate with.
• Model settings — Each model has its own settings panel (resolution, aspect ratio, output format, etc.).
• Generate button — Starts generation with all selected models simultaneously.

Right Column: Input & Results

• Input image — The image you want to edit, enhance, or use as a reference. Load it by dragging and dropping, pasting from clipboard, or browsing your computer.
• Results — Generated images appear here. Each result shows which model created it.

Core Tab

The Core tab is for general AI image generation and editing.

Selecting AI Models

The Core tab has 19 image models, newest first:

You can select multiple models at once. When you click Generate, all selected models run in parallel and you get results from each one.

Variant Count

Each active model shows a number badge (1–4). Click the number to cycle through variant counts. For example, setting Seedream 4.5 to 4 means it will run the same prompt 4 times, giving you 4 different results to choose from.

Model Settings

Each model has its own settings panel that appears when the model is selected:

• Resolution — Choose from presets (1K, 2K, 4K) or custom dimensions.
• Aspect ratio — Match input image, or choose from 1:1, 16:9, 9:16, 4:3, 3:4.
• Output format — PNG, JPEG, or WebP (varies by model).
• Provider — Nano Banana Pro and Nano Banana 2 run on Replicate, Kie.ai or Google; Seedream 4.5, Seedream 5 Pro and GPT Image 2 on Replicate or Kie.ai. If you only have a key for one of them, Rangy switches to it when you click Generate. A model button shows a dot when none of your keys can run it — hover it to see which provider it needs.
• Additional settings — Model-specific options like Go Fast mode, safety tolerance, prompt upsampling, etc.

Generation Workflow

Working with Results

• Full screen — Click the expand icon on any result to view it full screen. Zoom in (scroll wheel or double-click for 200%), pan around, and navigate between images.
• View prompt — In the full-screen viewer, click the Prompt button to see the prompt that was used to generate that image.
• Send to Affinity — Click the Affinity button to open the image directly in Affinity.
• Favorite — Star images you like.
• Copy — Single-click any image to copy it to clipboard.
• Delete — Moves images to trash (recoverable for 30 days).

Compare Mode

Compare Mode lets you view up to three images side by side in the full-screen viewer. It's ideal for comparing results from different AI models, before/after edits, or different prompt variations.

How to enter Compare Mode

Compare Mode controls

• Prompt display — The prompt of the first (center) image is always shown at the bottom of the compare view so you can reference it at any time.
• Synchronized zoom & pan — By default, scrolling and panning affects all panels equally, so you compare the exact same area across all images.
• Independent zoom & pan — Click the Unlock button to control each panel's zoom and position separately.
• Swap — Click the swap button that appears on the border between two panels to switch their positions. Useful for reordering images without leaving compare mode.
• Match size — Click the match-size button between the first two panels to make both images display at the same visual size, regardless of their original resolution.

Using with Affinity

Video Tab

The Video tab lets you generate AI videos from text prompts, images, or existing footage. You have access to 16 AI video models with full control over resolution, duration, aspect ratio, and more.

Video AI Models

How to Generate a Video

Kling 3 Motion Control

Motion Control is a special mode inside Kling 3 that transfers movement from a reference video onto a character image while preserving facial identity and expressions.

• Select Motion in Kling 3's Mode setting
• Add a reference image (required) — the character to animate
• Add a reference video (required) — the motion source
• Choose Character Orientation: Video (recommended) or Image
• Set resolution: 720p or 1080p

WAN 2.7 Video Edit

WAN 2.7 includes a Video Edit mode that lets you upload an existing video and transform it with a text prompt.

• Select Video Edit in WAN 2.7's Mode setting
• Add a video reference (required) — the video to edit
• Optionally add a reference image for style guidance
• Set Audio to Auto (AI decides) or Original (keep source audio)
• Duration 0 means use the full input video length; otherwise clips from 0 to the specified duration

Gemini Omni — Voice & Character Library

Gemini Omni is Google's multimodal video model. Beyond a text prompt it can take reference images and a video input, and it builds richer, controllable results using a reusable library of voices and characters.

• Create voice — open the Create Voice modal from the Gemini Omni settings card, pick one of 30 preset voices (preview each with a play button), name it, and optionally add a description. Voices are saved locally and can be attached to any future generation.
• Create character — give a reference image and (optionally) a voice to build a reusable character. You can combine up to three characters in one shot so the model keeps their look consistent across shots.
• Input quota: reference images + a video (counts as 2) + characters must total 7 inputs or fewer; the picker stops you at three characters per shot.

Viewing Video Results

Generated videos appear in the Results panel on the right. You can:

• Hover to preview, click to open in the fullscreen video player with custom controls
• Navigate between videos using arrow keys or on-screen buttons
• Use keyboard shortcuts: Space (play/pause), M (mute), F (fullscreen), Shift+Arrow (seek)
• Videos are saved locally to your Videos folder and appear in the History tab under the Videos section

Rangy Agent

The Agent is a chat-based way to create. Describe what you want in plain language and it picks the right model or tool, writes a detailed prompt, runs the generation, and drops every result onto a freeform canvas. In v2.6 the Agent's brain was upgraded to v2 — noticeably smarter at understanding requests, choosing models, and writing prompts, with access to every tool in Rangy (image, video, upscaling, and YouTube thumbnails).

The Agent runs on a server-hosted brain, so it needs an internet connection and an active license. It uses your own provider keys (Replicate / Kie / Google) for the actual generations, so you stay in control of cost.

The chat itself runs on your own AI key: with an Anthropic key it uses Claude Haiku 5.5 (the default) or Claude Sonnet 5.5; with a Google AI key, Gemini 3.5 Flash-Lite (the default) or Gemini 3.8 Flash. Switch engines from the chat header. The Agent can use every model in Rangy, including Ideogram 4.5, Nano Banana 2.1, Flux 3, GPT Image 2.5 and Vidu Q4.

Quick & Project Modes

• Quick mode — one-shot generations. Type something like “make a poster of a mountain at sunset” and it generates right away.
• Project mode — give a multi-step brief (a campaign, storyboard, or multi-shot ad) and the Agent returns a structured plan to approve, with estimated cost, before anything runs.

Canvas & Projects

• Freeform canvas — pan, zoom, and arrange every result on an infinite canvas. Group by model, date, or aspect ratio, and hide thumbs-down results.
• Projects — organize work into separate projects, each with its own chat history and canvas.
• Rate results — give any generation a thumbs-up or thumbs-down so the Agent learns your preferences over time.
• Per-card tools — iterate, make variations, upscale, remove background, or use a result as the reference for your next message, straight from each card.

Brand Library

Each project has a brand library: drop in reference images (logos, product shots, style references) and recall them by name in any message. The Agent passes them to the model so your brand stays consistent across generations.

AI Assistants (MCP)

Rangy includes a local MCP server, so AI assistants on your computer — Claude Desktop, Claude Code, Codex CLI, Gemini CLI, Cursor and other MCP apps — can write the prompt, pick the model and settings, and generate images and videos through Rangy. Everything they create lands in your Rangy library as usual. Requires a Pro license.

Pro Tab

The Pro tab provides advanced features for professional workflows.

Fast Actions (Image Restoration)

Fast Actions restore and enhance damaged or old images in one click — no prompt needed. They come in four model groups: GPT Image 2.5, Nano Banana Pro, Nano Banana 2 and Seedream 5 Pro, each with the same four actions.

You can select multiple actions within one group — for example Cracks + Restore + Colorize + Retouch — and they run as a single generation. Picking an action from a different group starts a new selection.

How to use:

Enhancement / Upscaling

Rangy offers four enhancement tools — Pruna, Crystal, Magnific and Skin — plus the YouTube Thumbnail maker:

Pruna Upscaler

• Provider: Replicate
• Scale factors: 2x, 4x, 6x, 8x (output is capped at 128 megapixels — Rangy rejects a factor that would go over before anything is spent)
• Price: by output size, from $0.005 (up to 4 MP) to $0.12 (up to 128 MP)
• Options: Enhance Realism (recommended for AI-generated images), Enhance Details, PNG or JPG
• Best for: The fastest, cheapest upscales

Crystal Upscaler

• Provider: Replicate
• Max output: 200 megapixels
• Scale factors: 2x, 4x, 6x, 8x, 10x
• Speed: 30–60 seconds typically
• Best for: General upscaling with excellent detail enhancement

Magnific

• Provider: Magnific API (formerly Freepik API)
• Best for: Premium upscaling with extensive control over the result
• Settings: Multiple parameters for fine-tuning

Skin

• Provider: Magnific API (formerly Freepik API)
• Best for: Skin texture rebuilding and portrait enhancement
• How it works: Takes over-saturated, over-retouched images and rebuilds realistic skin texture with natural detail

YouTube Thumbnail Maker

Turn any image into a polished, click-worthy 16:9 YouTube thumbnail with GPT Image 2. Select YouTube in the Enhancement section, load a reference image, set the options below, then click Generate.

• Reference image (required) — the base image to turn into a thumbnail.
• Title — turn it on to add a title. Type the exact text and place it on a 3×3 position grid, or leave the text empty to enhance existing on-image text or let AI pick a fitting title.
• Icon / logos — turn it on to overlay one or more logos. Add them by upload or paste, and place each one independently on its own 3×3 grid. Leave it empty if the icon is already in your image.
• Variations — generate 1–4 at once, and choose how different they should be (Slight = the same prompt each time; Very different = each variant gets a distinct style).
• Resolution — 1K, 2K, or 4K, with a live price estimate.
• Custom prompt (optional) — extra instructions that take priority over the defaults.

AI Prompt Studio

AI Prompt Studio helps you write, refine, and enhance your prompts using AI. It is powered by two AI models: Claude Sonnet 5.5 (Anthropic) and Kimi K3 (Moonshot AI).

• Write your prompt or describe what you want in natural language
• Select an action — improve, rewrite, make cinematic, simplify, add detail, etc.
• Choose your AI model — Sonnet 5.5 (fast, under $0.01 per prompt) or Kimi K3 (a reasoning model, about $0.01 per prompt)
• Preview the result and use it directly or continue editing

AI Prompt Studio also includes Color Grading presets for precise visual direction — apply cinematic looks, warm tones, cool palettes, and more to guide your generations.

JSON Extractor

The JSON Extractor analyzes any image and produces a structured JSON-style prompt capturing all visual properties — lighting, colors, composition, materials, mood, and more. This is useful for creating highly detailed, reproducible prompts.

Prompt Library

The Pro tab includes an expanded library of pre-built prompts and styles (9 new presets in v2.0) you can apply to your images:

• Run — Applies the style directly using Seedream 4.5.
• Modify — Loads the prompt into the Core tab so you can select a different model or adjust settings before generating.

Prompt Extractor

The Prompt Extractor uses Claude Sonnet 5.5 or Kimi K3 to analyze an image and extract a reusable prompt from it. This is one of Rangy's most powerful features.

How It Works

Portrait Group

Architecture Group

Image Group

Practical Examples

• Extracting portrait style: Take a celebrity portrait with distinctive lighting. Extract the style, then apply it to your own portrait using Nano Banana Pro. The result will have matching lighting, shadows, background, and visual feel.
• Extracting clothing: Find an image with a specific outfit. Extract the clothing prompt. Apply it to your own image. The generated result will show your subject wearing nearly identical clothing.
• Combining style and clothing: Extract clothing from one image and style from another, combine the prompts, and apply to your own photo.
• Architecture exterior finish: Have a building under construction? Find a finished building with the style you want. Extract the exterior finish prompt, then apply it to your building image.

Crop & Resize Tools

Crop Tool

• Upload, paste, or load an image from history.
• Freeform crop or select a specific aspect ratio.
• Save as a new image or replace the original.

Resize Tool

• Upload, paste, or load an image from history.
• Set your target dimensions.
• Save as new or replace the original.
• Useful for reducing image size before upscaling to control costs.

AI Providers & API Setup

Rangy connects to AI platforms with your own API keys, so you pay the provider directly for what you generate. The built-in setup guide — it opens on first launch, and anytime from Settings > API Keys — shows which key you need for images, videos, upscaling or the Agent, and tests each key the moment you add it (for Kie.ai it also shows your remaining credit). Most people start with Kie.ai.

Kie.ai (Recommended — start here)

Kie.ai runs most image models — including Nano Banana 2.1, Seedream 5 Flash, GPT Image 2.5, Qwen Image 3 and Grok Imagine 2.0 — every video model except P-Video, and the Sonnet 5.5 and Kimi K3 prompt tools. For models that also run on Replicate, Kie.ai is usually significantly cheaper and often faster.

Cost Comparison (Nano Banana Pro)

Speed: Kie typically generates 4K images in 15–50 seconds. On Replicate, the same generation can take 3–5 minutes.

Setup

Replicate

Replicate runs Flux 3 and the other Flux models, Ideogram 4.5, GPT Image 1.5, Seedream 5 Lite, the Crystal and Pruna upscalers and P-Video, and is a second route for Seedream 4.5, Seedream 5 Pro, Nano Banana Pro, Nano Banana 2 and GPT Image 2. It also keeps Kimi working in the prompt tools when you don’t have a Kie.ai key.

Setup

Google AI (Optional)

A Google AI Studio key runs Nano Banana Pro and Nano Banana 2 straight from Google, and powers the Agent on Gemini. See the Google AI key guide.

Anthropic (Optional)

An Anthropic key runs the Agent on Claude Haiku 5.5 or Sonnet 5.5, and the Sonnet 5.5 prompt tools when you don’t have a Kie.ai key. See the Anthropic key guide.

Magnific (For Magnific & Skin only)

The Magnific API — Freepik’s API under its new name — is needed only for the Magnific upscaler and the Skin Enhancer. Existing Freepik API keys keep working.

Setup

Billing: Magnific API usage is billed through your Magnific API plan — see magnific.com/api for current pricing.

Settings

Access settings from the gear icon in the app.

• License — Enter your license key to activate the app. You receive this key by email after purchasing.
• API Keys — Add keys for Kie.ai (recommended — most image models and almost every video model), Replicate (Flux, Ideogram 4.5 and the upscalers), Google AI (Nano Banana Pro/2 direct, Gemini for the Agent), Anthropic (Claude for the Agent) and Magnific (Magnific & Skin). Open the setup guide here to see which key you need.
• AI Assistants — Connect Claude, Codex, Gemini CLI and other MCP apps to Rangy (Pro). See AI Assistants (MCP).
• Storage — Set the folder where Rangy stores your generated images. You can change this to any folder on your computer.
• Appearance — Change the app's theme (light/dark mode).

Cost Reference

Generation Models

Crystal Upscaler (Replicate)

Pricing is based on output megapixels — the larger the final image, the higher the cost. Rangy shows the output size and price before you upscale:

Video Generation Models

All on Kie.ai except P-Video (Replicate).

Video costs are calculated as: cost per second × duration. For example, a 10-second Kling 3 Standard video without audio costs approximately $0.70.

Prompt Extraction & Other

Tips & Best Practices

• Start with Kie.ai — It runs most models and is usually the cheaper, faster route for models that also run on Replicate.
• Draft cheaply, finish in quality — Seedream 5 Flash ($0.016) and Nano Banana 2.1 ($0.02) are great for exploring ideas; switch to a premium model for the final.
• Text in the image? Try Ideogram 4.5 — It’s known for accurate lettering on logos, posters and signs.
• Real places or products? Use Flux 3 with Web Research — It looks them up before generating.
• Generate multiple variants — Set the variant count to 3–4 to get several options at once. Pick the best result.
• Use Nano Banana Pro for accuracy — When you need the most precise image editing (especially with extracted prompts), Nano Banana Pro gives the best results.
• Resize before upscaling — To reduce costs, first resize your image smaller using the Resize tool, then upscale with Crystal.
• Save prompts as favorites — When you get a good extracted prompt, favorite it. Reuse it across many images for a consistent look.
• Match aspect ratio to input — Use "Match Input Image" aspect ratio when editing existing images to avoid size mismatches.
• Failed generations cost nothing — If a generation fails on any provider, you won't be charged. Just retry.
• Keep Replicate credit above $10 — This removes rate limits and gives you a higher generation quota per minute.

Troubleshooting

Generation fails repeatedly

• Replicate: Their servers may be temporarily busy. Wait 10–20 minutes and retry. Issues typically resolve within an hour.
• Kie: Servers may need to warm up. Retry 2–3 times. After that, generations will succeed reliably.

A model button shows a dot

That model needs an API key you haven’t added yet. Hover the button to see which provider runs it, then add that key in Settings > API Keys (the setup guide there shows the steps).

Rate limiting

If you have less than $10 credit on Replicate, there's a limit on how many images you can generate per minute. Add more credit to increase the limit.

Image not positioned correctly

AI models may not always match the exact composition of your input. Generate multiple variants (set count to 3–4) and pick the best result.

Result image is larger/smaller than expected

Some models may output images at different sizes than the input. Try a different model — Nano Banana Pro is typically the most accurate at maintaining input dimensions.

Blue screen on Windows installation

This is the Windows SmartScreen warning. Click "More info" then "Run anyway." It is safe and normal — it will stop appearing once the developer account is fully verified with Microsoft.