Agent Caption — effect 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.

Local captions in any language: Whisper words, forced-aligned timing, SRT/VTT, burned styles or a GSAP captions.js.

What it does

A Python toolkit an agent drives from SKILL.md: Whisper large-v3 (faster-whisper) writes the words, WhisperX or Meta's MMS_FA aligner times every word on the waveform, and caption.py burns one of 15 styles with ffmpeg (falling back to Pillow overlays when ffmpeg lacks libass) or just writes .srt/.vtt. scripts/caption.py emits a captions.js layer (word, line or karaoke presets such as hormozi, pill, neon) that mounts on a GSAP timeline in an HTML composition, and Argos Translate makes offline translated subtitle tracks. Extras include --script alignment of known lyrics, Demucs vocal isolation for songs, a glossary prompt and a deterministic QA report. Setup is heavy (torch, WhisperX and about 4 GB of models) and Hinglish/code-switched speech is its focus.

Use when

Not for

Install

Models it drives: whisper-large-v3, whisperx, mms-fa, htdemucs

Example output

https://skills.varg.ai/previews/agent-caption.mp4 — The repo's demo video: the creator talks to camera while captions generated locally by the skill switch language line by line (English, then Spanish), with the language tag on top. First 8 s of 58.5 s. (credit: Ali Hamza Kamboh (agent-caption), https://github.com/ahkamboh/agent-caption#readme)

Used for

More captions & subtitles skills

Run the generation on varg

varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs.