Video caption generator skill

name: video-caption-generator

by ericosiu·MIT license·★ 3,615 Stars on the repo·GitHub ↗

Use now

ericosiu/main1 file shown

SKILL.md2.9 KB

Browse every file on GitHub↗

SKILL.md

Show the full text86 lines

video-caption-generator/SKILL.md86 lines · 2.9 KB

PreviewCode

Outline

RawView on GitHub

Video Caption Generator

Processes new MP4s from a Google Drive folder: transcribes, deduplicates, generates captions + titles.

Drive Folder Setup

Configure your Google Drive folder IDs before first use:

Folder	Purpose
Main / To Schedule	New clips are dropped here for processing
Scheduled	Already posted/scheduled clips (moved after publishing)
A/B	Title variants for later posting

Set your folder IDs in folder-map.json or pass them directly via --folder-id.

Quick Run

python3 skills/video-caption-generator/scripts/process_videos.py \
  --folder-id YOUR_FOLDER_ID

Processed video IDs are logged to processed_ids.json so already-seen videos are skipped on future runs.

A/B Variant Handling

Videos with identical transcripts but different filenames (e.g., 0411.mp4, 0411(1).mp4) are A/B title variants — same audio, different on-screen title. The script processes ALL variants (no dedup skipping) and tags them as A/B variants in the output.

Output Format

For each unique new clip, output:

*<filename>*
📝 *Transcript:* <raw spoken words>
🎬 *Caption:* <social-friendly 2-4 sentence caption>
📺 *YT/FB Title:* <punchy title under 60 chars>

Caption + Title Style Guide

Caption: First person, conversational, no hashtags, 2-4 sentences. Hook first, insight second.
Title: Curiosity-driven, under 60 chars, no "How I..." unless earned. Lead with the tension or number.

Dependencies

whisper (local installation, model: turbo)
Google Drive CLI or SDK for listing/downloading files
Anthropic API key (set via ANTHROPIC_API_KEY environment variable)

Configuration

Set ANTHROPIC_API_KEY in your environment
Update GWS_GATEWAY in scripts/process_videos.py to point to your Google Drive CLI
Update WHISPER_BIN if Whisper is installed at a different path

Adding New Folders

Pass --folder-id <ID> for different folders. Each folder uses the same shared processed_ids.json log (Drive IDs are globally unique so there's no collision).

1	---
2
3	## Preamble (runs on skill start)
4
5	```bash
6	# Version check (silent if up to date)
7	python3 telemetry/version_check.py 2>/dev/null \|\| true
8
9	# Telemetry opt-in (first run only, then remembers your choice)
10	python3 telemetry/telemetry_init.py 2>/dev/null \|\| true
11	```
12
13	> Privacy: This skill logs usage locally to `~/.ai-marketing-skills/analytics/`. Remote telemetry is opt-in only. No code, file paths, or repo content is ever collected. See `telemetry/README.md`.
14
15	---
16	name: video-caption-generator
17	description: >
18	Transcribe short videos from a Google Drive folder, deduplicate by content,
19	and generate social captions + YouTube/Facebook titles. Use when new video
20	clips are dropped into a Drive folder and you need transcript, caption, and
21	title for each unique clip.
22	Trigger phrases: "process videos in drive", "transcribe new clips",
23	"caption these videos", "generate titles for these clips".
24	---
25
26	# Video Caption Generator
27
28	Processes new MP4s from a Google Drive folder: transcribes, deduplicates, generates captions + titles.
29
30	## Drive Folder Setup
31
32	Configure your Google Drive folder IDs before first use:
33
34	\| Folder \| Purpose \|
35	\|--------\|---------\|
36	\| Main / To Schedule \| New clips are dropped here for processing \|
37	\| Scheduled \| Already posted/scheduled clips (moved after publishing) \|
38	\| A/B \| Title variants for later posting \|
39
40	Set your folder IDs in `folder-map.json` or pass them directly via `--folder-id`.
41
42	## Quick Run
43
44	```bash
45	python3 skills/video-caption-generator/scripts/process_videos.py \
46	--folder-id YOUR_FOLDER_ID
47	```
48
49	Processed video IDs are logged to `processed_ids.json` so already-seen videos are skipped on future runs.
50
51	## A/B Variant Handling
52
53	Videos with identical transcripts but different filenames (e.g., `0411.mp4`, `0411(1).mp4`) are A/B title variants — same audio, different on-screen title. The script processes ALL variants (no dedup skipping) and tags them as A/B variants in the output.
54
55	## Output Format
56
57	For each unique new clip, output:
58
59	```
60	<filename>
61	📝 Transcript: <raw spoken words>
62	🎬 Caption: <social-friendly 2-4 sentence caption>
63	📺 YT/FB Title: <punchy title under 60 chars>
64	```
65
66	## Caption + Title Style Guide
67
68	- Caption: First person, conversational, no hashtags, 2-4 sentences. Hook first, insight second.
69	- Title: Curiosity-driven, under 60 chars, no "How I..." unless earned. Lead with the tension or number.
70
71	## Dependencies
72
73	- `whisper` (local installation, model: turbo)
74	- Google Drive CLI or SDK for listing/downloading files
75	- Anthropic API key (set via `ANTHROPIC_API_KEY` environment variable)
76
77	## Configuration
78
79	1. Set `ANTHROPIC_API_KEY` in your environment
80	2. Update `GWS_GATEWAY` in `scripts/process_videos.py` to point to your Google Drive CLI
81	3. Update `WHISPER_BIN` if Whisper is installed at a different path
82
83	## Adding New Folders
84
85	Pass `--folder-id <ID>` for different folders. Each folder uses the same shared `processed_ids.json` log (Drive IDs are globally unique so there's no collision).
86