Scriptify

Scriptify is an AI writing tool designed for authors and creatives. It uses quantized Whisper models with ONNX Runtime for near-real-time, on-device transcription, helping you capture your thoughts as you speak without sending audio to a transcription service.

Visit our website for screenshots/minimum requirements, and supported platforms: https://infinitode.netlify.app/apps/scriptify.

Features in v1.2.1-beta:

• Near Real-time Transcription: Convert spoken words into text almost instantly.
• Focused Writing Studio: A calmer editor with accessible formatting tools, and a dedicated reading mode with adjustable text size, reading progress, and a persistent light/dark page theme.
• Richer Formatting & Exporting: Use multiple heading levels, quotes, lists, and inline styles, then export your work in popular formats with structure preserved.
• Useful Writing Statistics: Track word and character counts, reading time, sentence and paragraph counts, vocabulary variety, and estimated readability.
• In-app Dictionary: Look up selected words quickly without leaving the editor.
• Ideas & Starting Thoughts: Browse hundreds of story prompts or get a first line to help move past a blank page. Add an opening thought to your draft with one click.
• Automatic Story Library: Drafts are saved to your per-user application data as you write, then appear in the sidebar and full-screen library the next time you open Scriptify. Browse start and edit dates, rename, copy, or delete stories.
• Autosave & Backup: The active story is saved automatically, the in-app close button waits for pending saves, and a recovery copy protects recent edits if the app exits unexpectedly. A separate desktop safety snapshot is also created every 10 minutes.
• No Limits: Scriptify is 100% free, with no limits or ads.

Current development

• Replaced the PyTorch/OpenAI Whisper runtime with ONNX Runtime and int8 Whisper graphs. Model weights download on demand and stay outside the application bundle.
• Simplified the model chooser and made its download status honest: the downloader does not expose byte-level progress, so the UI uses an indeterminate indicator instead of fabricated percentages.
• Added a saved light/dark page preference to reading mode.
• Fixed a bug where auto-saved stories were lost on every app restart. pywebview 6.x defaults to private_mode=True, which wipes localStorage (and all WebView2 browser storage) when the app exits. webview.start() is now called with private_mode=False and a stable storage_path pointing to %APPDATA%/Scriptify/webview-storage (platform-equivalent on macOS/Linux), so stories saved to localStorage survive between launches.
• Kept the existing editor, story library, backups, formatting, exports, writing statistics, dictionary, and microphone workflow.

Changelog v1.2.0-beta:

• Refreshed the writing studio with reading mode, writing metrics, formatting tools, and inspiration prompts.
• Added an automatically saved story library with desktop application-data storage and a browser local-storage fallback.

Changelog v1.0.0-beta:

• Created a launch window for model selection and downloads.
• Modified transcription logic to allow for "real-time" audio transcription by recording the audio in chunks and then processing it in the background.
• Introduced windowed modes for all platforms (no console windows).
• Updated UI with animations and new shortcuts.
• A ton of bug fixes, including fixing ffmpeg extraction, openai-whisper hidden imports and files, and file path fixes.
• Several export formats have been added, though there are some issues with <h2> headings.
• Added an information panel for writing statistics.

Important

Whisper model files are downloaded separately on first use and are not embedded in the release executable. The build bundles ONNX Runtime for CPU inference; final artifact sizes should be checked from the platform build workflow.

Note

Releases for Scriptify will not be included in Distributables since the release binaries are already distributed here.

Installation (from source)

1. Clone the repository:

   git clone https://github.com/Infinitode/Scriptify.git

2. Install dependencies:

   pip install -r requirements.txt

3. Run Scriptify:

   python main.py

Installation (releases)

1. Download the release binary for your platform: Head to the releases. Select the latest release. Download the latest release binary for your platform.

2. Run Scriptify: Run the Scriptify application file.

Usage

• On first launch, choose one of the six Whisper sizes. Scriptify downloads only the selected int8 ONNX model into models/<model-name>/ beside the application; larger models take more storage and memory.
• The first download requires an internet connection and may take a while. Cached models are loaded locally on later launches, and microphone audio is processed on-device.
• The writing studio opens after the selected model is ready. You can write normally or dictate into the current draft; use the reading-mode controls to select a light or dark page.
• Your current story is saved automatically. Use Recent stories in the sidebar to resume, rename, duplicate, or remove drafts; All stories opens the full library.
• Use Export or Save draft when you want a separate file to share or archive. Scriptify also creates a desktop safety backup every 10 minutes.

Automatic story files are kept in Scriptify's per-user application data folder:

• Windows: %APPDATA%/Scriptify/stories
• macOS: ~/Library/Application Support/Scriptify/stories
• Linux: $XDG_DATA_HOME/Scriptify/stories, or ~/.local/share/Scriptify/stories

WebView2's persistent browser profile (used to preserve localStorage between sessions) is stored alongside:

• Windows: %APPDATA%/Scriptify/webview-storage
• macOS: ~/Library/Application Support/Scriptify/webview-storage
• Linux: $XDG_DATA_HOME/Scriptify/webview-storage, or ~/.local/share/Scriptify/webview-storage

When running in a regular browser rather than the desktop app, the library uses that browser's local storage for the current site.

Minimum Requirements

• A supported desktop operating system and audio input device.
• An internet connection for the initial model download; cached models can be used offline afterward.
• At least 2 GB of RAM for the smallest model. More RAM and disk space are recommended for medium, large, and turbo models.
• A CPU with AVX2 is recommended on x86-64 systems; a GPU is optional. The bundled runtime uses ONNX Runtime's CPU execution provider by default.

Tip

If you are unsure which model to choose, start with Tiny or Base. You can switch models on a later launch; each selected model is cached separately in models/.

License & Redistribution

This project is open source. However, it may not be redistributed without significant changes or under the same name. Please refer to the LICENSE file for more details on the terms of use.

Scriptify bundles a few third-party assets so the interface renders without an internet connection:

• Bootstrap Icons — MIT licence, in web/vendor/bootstrap-icons/.
• Inter and Literata variable fonts — SIL Open Font License 1.1, in web/fonts/.

Each asset's licence is kept alongside the files it covers.

Contributing

Contributions are welcome! Feel free to open issues or submit pull requests to help improve Scriptify. You can read the CONTRIBUTING file for more information and general contribution guidelines.

Happy writing!