vexy-lines-run

Desktop GUI for Vexy Lines style transfer: load images, .lines files, or video, pick a style, export.

Built with CustomTkinter. Runs on macOS, Windows, and Linux wherever Tk is available.

Install

pip install vexy-lines-run

Core dependencies (CustomTkinter, tkinterdnd2, CTkMenuBarPlus, Pillow, loguru) install automatically. Video processing dependencies (OpenCV, PyAV) are pulled in transitively by vexy-lines-apy. To pin them explicitly:

pip install "vexy-lines-run[video]"

For AI-assisted layer rename:

pip install "vexy-lines-run[ai]"

Launch

vexy-lines-run

Or:

python -m vexy_lines_run

From Python:

from vexy_lines_run import launch
launch()

Features

Three input tabs

• Lines: load .lines files; export embedded previews or apply a new style
• Images: load PNG, JPG, WEBP, and other rasters; the app applies the style via the MCP API
• Video: load MP4, MOV, MKV, or similar; per-frame style transfer with audio passthrough

Style picker

Select a primary style from any .lines file. Optionally select an end style: the app interpolates the two linearly across the input sequence. Both show inline thumbnail previews. The app preserves matching image-filter chains on fills and interpolates them as part of the style.

AI Rename Layers & Fills

Lines ▸ AI Rename Layers & Fills… renames a .lines document's generic captions (Layer, Blended, Linear) to describe what each fill actually draws, then names each layer from its fills. Only the captions change. After a confirmation and a Save dialog, each fill is rendered in Vexy Lines and described by a vision model on a background thread (progress shows in the title bar); the renamed copy defaults to <stem>-renamed.lines. Needs the [ai] extra (pip install "vexy-lines-run[ai]"), the Vexy Lines app, and an OpenAI-compatible /v1 LLM endpoint. Set it in Lines ▸ AI Rename Settings… or via VEXY_LINES_LLM_API_URL, VEXY_LINES_LLM_API_KEY, VEXY_LINES_LLM_MODEL_VISION, and VEXY_LINES_VLM_MODEL. See the full guide.

Export formats

Drag-and-drop onto any input list (via tkinterdnd2)

Background processing: export runs on a daemon thread; the progress bar updates live and the UI stays responsive

Architecture

app.py          App(CTk)         — window, three tabs, style pickers, export bar
processing.py   process_export() — background thread dispatcher for lines/images/video
video.py        probe()          — PyAV-based video metadata and per-frame processing
widgets.py      CTkRangeSlider   — dual-handle range slider for video frame selection

Style transfer calls into vexy-lines-apy (MCPClient, apply_style, interpolate_style). Video uses PyAV for mux/demux and OpenCV for frame extraction.

Job folders (crash-safe exports)

All exports create a persistent job folder alongside the output directory. Every intermediate artifact (.lines documents, .svg exports, rasterized frames) is saved there. If the app quits or the process is interrupted mid-export, re-running the export resumes from where it left off.

The GUI never deletes job folders automatically. Use the CLI with --force to start fresh or --cleanup to remove the folder after completion.

Full documentation

Read the docs for the complete GUI guide, API reference, and more examples.

License

MIT