How to use Claude to build Lua widgets for mpv
The author reports a GUI library and several working media scripts produced over a weekend, validated through demos while acknowledging likely undiscovered bugs.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Build [Lua widget library] for mpv using libASS subtitle drawing primitives. Include [required widgets] and an interactive demo. After feedback, use it for [media script]. Check mpv APIs and the first-party OSC when implementing scrolling, overlays and drag interactions.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- mikkupikku
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports a GUI library and several working media scripts produced over a weekend, validated through demos while acknowledging likely undiscovered bugs.
- Human involvement
- The author tested demos, read a small portion of the code and corrected API assumptions; they expected undiscovered bugs.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Generate a Lua widget library from subtitle drawing primitives.
- Test the evolving demo during library development.
- Build media scripts using the generated GUI library.
The author used Claude to generate roughly 10,000 lines of Lua for a GUI library built from libASS subtitle drawing primitives. The library included nearly two dozen widgets, with nested scrollable containers, clipping, dropdown menus, and animated tab bars. The author tested the evolving demo rather than reading the entire implementation, estimating they read about 100 lines that day.
The following day, they had Claude create several projects using the library. These included a video looping script, graphical updates to older text-only scripts, a playlist visualizer supporting drag reordering, and controls for text-to-speech reading of subtitles. Another project let viewers choose and reorder clips and export an edit decision list.
The author's familiarity with mpv guided corrections. They directed Claude to use WHEEL_LEFT and WHEEL_RIGHT for horizontal scrolling, rejected a proposed PNG overlay because that path required raw pixel data, and sent it to inspect the first-party OSC implementation for dragging behavior. They describe working demos and rapid progress, while openly expecting undiscovered bugs. The post does not establish comprehensive validation of the library.
Who it's for: Experienced mpv users building personal desktop scripts who can test interactions and correct API misunderstandings.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · mikkupikku
Frequently asked questions
How did the author validate the generated code?
They tested the demo Claude updated during development and read about 100 lines, rather than reviewing the entire roughly 10,000-line library.
Which mpv assumptions needed correction?
The author corrected horizontal scrolling to use WHEEL_LEFT and WHEEL_RIGHT, explained the raw-pixel overlay requirement, and directed Claude to inspect the first-party OSC for dragging behavior.
More details
Remaining steps
- Correct scrolling and overlay assumptions using mpv knowledge.
- Inspect the first-party OSC to investigate dragging behavior.
About this account
Claude generated the library, built projects on top of it and investigated existing OSC code after the author directed it there.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47307211. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: mikkupikku on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude · Lua · libASS · mpv
Lua · mpv · GUI
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.