Caption Pipeline Libraries — guide skill for captions & subtitles

For AI agents: find the right skill in one call — curl -s "https://skills.varg.ai/api/search?q=<what the person wants>" returns the matching use case and the best skills with install commands, as markdown (Russian works too). Full index: https://skills.varg.ai/llms.txt. Every page has a markdown twin: add .md to its path. Before generating anything, show the person the options on the use-case page.

Open-source libraries for every caption step: word timing, alignment, SRT/VTT/ASS, in-page or ffmpeg rendering.

What it does

A pointer pack for the caption half of a code-to-video pipeline, in the order a render runs: transcribe with word timestamps (whisper.cpp, faster-whisper, WhisperX), tighten timing against a known script (ctc-forced-aligner, Montreal Forced Aligner), read and write SRT, WebVTT and ASS (@remotion/captions, subtitle.js, subsrt-ts, ass-compiler, pysubs2), then render the captions in the page (HyperFrames caption components, ASS.js, JASSUB) or burn them in with ffmpeg's libass filters. Licenses, stars and last commits were checked on GitHub on 2026-10-26, and each note flags its trap: CC-BY-NC alignment weights, Remotion's company license on its whisper installer, LGPL/GPL native code inside JASSUB's npm package, FFmpeg build flags. Draft until the pack is published in vargHQ/skills with a manifest.

Use when

Not for

Install

Not published as a standalone skill yet. Source: https://github.com/vargHQ/skills

Example output

https://skills.varg.ai/previews/craft-caption-libraries.mp4 — Word-timed caption burn, the end of the pack's pipeline, no model run: the film's official English subtitles (TOS-en.srt, lines 17-20, from the Blender mirror) split into pages of up to four words, word onsets set by hand from the audio spectrogram (no ASR run), each page rendered to transparent PNGs in headless Chrome with Montserrat ExtraBold (OFL; active word in an orange box, next words dimmed) and composited with ffmpeg overlay=enable='between(t,a,b)' (this ffmpeg build has no libass). Footage: Tears of Steel 720p (CC BY 3.0, (CC) Blender Foundation | mango.blender.org), 03:16.9-03:26.6, cropped to 16:10; the preview is silent. (credit: varg captions on Tears of Steel (CC BY 3.0, (CC) Blender Foundation | mango.blender.org), https://mango.blender.org/)

More captions & subtitles skills