Image Prompt Library

English · 繁體中文 · 简体中文

Image Prompt Library is a local-first app for saving images with their prompts, sources, and notes. Browse visually, organize cards into collections and tags, and find a prompt when you want to use it again.

Your library uses local SQLite and local image files. You do not need an account to save, search, or organize it. Optional ChatGPT and Grok connections let you generate images and suggest titles; those requests run on the provider's servers, not on your computer.

The local app with optional sample references loaded. A new library starts empty.

v0.11.0 is the current stable release, available from GitHub Latest. It adds Grok image generation and title suggestions, multi-image cards, and searchable batch Tag and Move controls. See the v0.11.0 release notes for the full changes.

Requires Windows 10/11, PowerShell 5.1+, and Python 3.10+. Install Python first; the installer does not install it for you.

irm https://raw.githubusercontent.com/EddieTYP/image-prompt-library/main/scripts/install.ps1 | iex

The installer starts the app in the background and opens your browser. Use image-prompt-library stop to stop it.

Requires Python 3.10+ and curl. Release installs do not require Node.js.

curl -fsSL https://raw.githubusercontent.com/EddieTYP/image-prompt-library/main/scripts/install.sh | bash
image-prompt-library start

Keep the terminal open and visit http://127.0.0.1:8000. Press Ctrl-C in that terminal to stop the server.

For an inspect-before-running installation, updates, rollback, or uninstall, see the installation guide. If the app does not start, run image-prompt-library status and image-prompt-library doctor, then check Troubleshooting.

A fresh local library starts empty. Choose + Add, add an image and prompt, give the card a title, and save it. Collections, tags, notes, and source links are optional.

To browse examples first, import one of the optional sample bundles:

image-prompt-library sample-data en       # English collection names
image-prompt-library sample-data zh_hans  # Simplified Chinese collection names
image-prompt-library sample-data zh_hant  # Traditional Chinese collection names

Run only the command for your preferred collection language. Source titles, prompts, and available language variants are preserved. A larger Traditional Chinese pack is also available:

image-prompt-library sample-data zh_hant awesome-gpt-image-2

• Explore groups image references by collection; Library provides the full card view and editing controls.
• Keep original and translated prompts together. Select a language variant to read or copy it.
• Use keyword search across titles, prompts, tags, collections, sources, and notes, or combine it with structured search filters such as tag:portrait, collection:architecture, and sort:title.
• Select several cards to tag, move, favorite, archive, or delete them. Tag suggests existing tags and Move lets you search for a collection instead of typing its exact name.
• Keep multiple images in one card. In Edit, choose the cover image, reorder images, change result/reference roles, or remove one image without deleting the whole card.

Explore is for browsing collections; Library is for managing individual cards.

A card keeps its images, prompt variants, and source together. New generated images also retain their provider and model details.

Generation is optional and requires a local install and an eligible provider account. Grok OAuth is experimental. Account access, usage limits, and charges are controlled by the provider; connecting does not guarantee free or unlimited generation. No API key is required by these OAuth connections.

1. Open Config → Providers, choose ChatGPT / Codex OAuth or Grok OAuth, and complete the browser authorization. Return to the app and choose Check authorization if prompted.
2. Choose a Default AI provider. This is saved in the current browser. You can override it for a generation session without changing the default.
3. Open Generate, enter a prompt, and optionally add reference images. Reusable prompts can include {{variables}}, such as {{subject}}; fill them before generating.
4. Choose the available ratio, quality, and other controls. Generate one image or a set of 3, 5, or 10, then review completed results from the Work queue.
5. Use Save as new item, or attach a result to its unchanged source card. When reviewing a set, later results can join the card created from the first saved result, so the images stay together.

Grok uses grok-imagine-image-2.0, with Low/Medium quality, 1K/2K resolution, and up to three reference images. ChatGPT supports up to four references and has its own model and quality choices. Switching provider updates the available controls.

Choose a provider in the generation window. This is separate from the default saved in Config.

In Add, Edit, or a generated-result save form, choose Suggest title to request a short title from the prompt text. Review the suggestion and choose Use title to apply it; your current title is not overwritten automatically.

Add/Edit uses your default AI provider. A generated result uses the provider that made it, shown as via ChatGPT or via Grok. If the chosen provider is disconnected, the app asks you to reconnect; it does not silently send the prompt to another provider.

See the generation guide for connection steps, reference-image limits, result handling, and troubleshooting.

Open the demo to browse collections, search examples, and copy public sample prompts without installing anything.

The hosted demo uses public sample data. Editing, private-library management, and generation are available only in a local install.

The demo references and optional sample bundles retain their source links and license information. The sample prompts and artwork are not original Image Prompt Library content.

See sample-data/README.md for pack details and checksums.

• Keep your own private prompt/image library on your computer. There is no hosted user database or built-in cloud sync.
• Image generation sends the entered prompt and selected reference images to the provider you choose. Title suggestions send only prompt text—not your images, existing title, tags, or notes.
• Provider credentials are stored separately from the library and excluded from portable backups and sample exports.
• The app listens on 127.0.0.1 by default. Do not expose it to a network unless you understand the access risks.

• Installation and updates
• Image generation and title suggestions
• Backup and restore
• Troubleshooting
• Development setup and Contributing
• Roadmap

Application code is AGPL-3.0-or-later. See LICENSE and NOTICE. Sample data and third-party assets have separate licenses, listed above.

Commercial licenses are available by arrangement with the maintainer for use under terms other than the AGPL.