Short Generator Skill

Generate YouTube Shorts using video clips sourced from multiple social media platforms.

Quick Command Format

/short-generator <subject> [voice:<voice_id>] [hashtags:<tag1,tag2>] [sources:<platform1,platform2>] [ratio:<9:16|16:9|1:1|4:5|21:9>] [music:<on|off>] [subtitle_position:<pos>]

Any missing parameters will be collected interactively.

Workflow Steps

Step 1: Parse Input & Validate Environment

Parse the user's command for optional inline params:

• voice: — TTS voice ID (default: ask user)
• hashtags: — comma-separated search terms/tags (default: ask user)
• sources: — comma-separated source platforms (default: ask user)
• ratio: — aspect ratio (default: 9:16)
• music: — on or off (default: ask user)
• subtitle_position: — subtitle position (default: center,bottom)

Extract the subject (first positional arg after removing all key:value tokens).

Pre-flight check: Ensure the Flask backend is running on port 8080:

curl -s -o /dev/null -w "%{http_code}" http://localhost:8080/api/models

If not running, ask the user to start it (cd Backend && python main.py).

Fetch available voices and settings:

curl -s http://localhost:8080/api/models | python3 -m json.tool
curl -s http://localhost:8080/api/settings | python3 -m json.tool

Extract the available voices list and aspect ratios from the response for use in questions.

Step 2: Generate Script

Call the script generation endpoint:

curl -s -X POST http://localhost:8080/api/script \
  -H "Content-Type: application/json" \
  -d '{"videoSubject": "<subject>", "extraPrompt": "Make it engaging and viral", "aiModel": "g4f", "scriptTemplate": "viral_shorts"}'

Show the generated script to the user with clear formatting. Ask if they want to:

• Accept and continue
• Regenerate with a different prompt
• Edit the script manually (let them paste a custom script)

Store the final script and the search_terms from the response.

Step 3: Collect Missing Parameters

For each parameter the user didn't provide inline, ask:

1. Voice: Ask "Which TTS voice?" with available voices from Step 1. Show voice IDs.
2. Hashtags/Search Terms: Ask "What search terms or hashtags should I use?" (e.g., Messi goal, World Cup 2026, soccer highlights). Get at least one.
3. Source Platforms: Ask "Which platforms to search?" — offer platforms from the reference table below. Default: instagram,facebook,tiktok,youtube,bilibili,douyin. Let them pick multiple.
4. Aspect Ratio: Ask "Which aspect ratio?" — show options from settings (9:16, 16:9, 1:1, 4:5, 21:9). Default: 9:16.
5. Music: Ask "Add background music?" — on or off.
6. Subtitle Position: Ask "Subtitle position?" — options: center,bottom, center,center, center,top. Default: center,bottom.

Collect responses and store them.

Step 4: Find Video URLs Across Selected Platforms

Use the search terms from Step 3 to find at least 5 video URLs across the user's chosen platforms.

For each platform, use the appropriate search strategy:

Instagram

• Chrome DevTools: Navigate to https://www.instagram.com/explore/tags/<hashtag>/, take snapshot, look for /reel/ links
• Web search: site:instagram.com/reel <search_terms>

Facebook

• Chrome DevTools: Navigate to https://www.facebook.com/search/videos/?q=<search_terms>, look for video URLs
• Web search: site:facebook.com/watch <search_terms>

X / Twitter

• Chrome DevTools: Navigate to https://x.com/search?q=<search_terms>&src=typed_query&f=video, look for video post URLs
• Web search: site:x.com <search_terms> video

TikTok

• Chrome DevTools: Navigate to https://www.tiktok.com/search/video?q=<search_terms>, look for video URLs (contain /video/)
• Web search: site:tiktok.com <search_terms>

YouTube

• Chrome DevTools: Navigate to https://www.youtube.com/results?search_query=<search_terms>, look for /watch?v= links
• Web search: site:youtube.com <search_terms> — prefer /shorts/ URLs if available

Bilibili (Chinese)

• Chrome DevTools: Navigate to https://search.bilibili.com/all?keyword=<search_terms>, look for video URLs (contain /video/)
• Web search: site:bilibili.com <search_terms>

Douyin (Chinese TikTok)

• Chrome DevTools: Navigate to https://www.douyin.com/search/<search_terms>, look for video URLs
• Web search: site:douyin.com <search_terms>

Kuaishou (Chinese)

• Web search: site:kuaishou.com <search_terms>

Weibo (Chinese)

• Web search: site:weibo.com <search_terms> video

Nico Nico Douga (Japanese)

• Chrome DevTools: Navigate to https://www.nicovideo.jp/search/<search_terms>, look for /watch/ URLs
• Web search: site:nicovideo.jp <search_terms>

Fallback — Ask user: If automated search finds fewer than 5 URLs, ask the user to provide video URLs directly from any platform.

Collect all URLs into an array. Deduplicate. Ensure at least 5 total.

Step 5: Download Videos

For each video URL, download via the backend API:

curl -s -X POST http://localhost:8080/api/instagram/download \
  -H "Content-Type: application/json" \
  -d '{"url": "<video_url>"}'

The backend uses yt-dlp which supports all the platforms listed above. If a URL fails:

• Log the error and move to the next URL
• Try appending ? query params or alternate URL formats (e.g., short URL)
• As a last resort, skip that URL and note it to the user

Save each successful download's filename into a list. The files are stored at Backend/static/generated_videos/instagram/<filename>. Note: even though the directory is named instagram/, the backend's yt-dlp downloader handles all platforms.

Show progress: "Downloaded 3/5 videos..."

Construct the full paths:

import os
backend_dir = os.path.dirname("Backend")
video_paths = [os.path.join(backend_dir, "static", "generated_videos", "instagram", f) for f in filenames]

Step 6: Show Full Configuration Preview

Show the user a clear preview of everything configured:

╔══════════════════════════════════════════╗
║           VIDEO CONFIGURATION            ║
╠══════════════════════════════════════════╣
║ Subject:     <subject>                   ║
║ Voice:       <voice_id>                  ║
║ Aspect:      <ratio>                     ║
║ Music:       <on/off>                    ║
║ Subtitles:   <position>                  ║
║ Sources:     <platforms list>            ║
║ Clips:       <count> videos              ║
╠══════════════════════════════════════════╣
║ Script:                                  ║
║ <first 300 chars of script>...           ║
╚══════════════════════════════════════════╝

Ask: "Generate the video with this configuration?" (Yes/No)

Step 7: Generate Final Video

On confirmation, call the search-and-download endpoint with the downloaded video paths:

curl -s -X POST http://localhost:8080/api/search-and-download \
  -H "Content-Type: application/json" \
  -d '{
    "search": <search_terms>,
    "script": "<script>",
    "aiModel": "g4f",
    "voice": "<voice_id>",
    "subtitlesPosition": "<subtitle_position>",
    "aspectRatio": "<ratio>",
    "subtitleTemplate": "classic",
    "directVideoPaths": <video_paths>,
    "useMusic": <true/false>
  }'

This will use the downloaded videos directly (skipping Pexels search).

Step 8: Show Result

Show the final video details:

✅ Video generated!
   Path: /static/generated_videos/<filename>.mp4
   Open: http://localhost:8080/static/generated_videos/<filename>.mp4

Ask if they want to add background music (if not added yet) via:

curl -s -X POST http://localhost:8080/api/addAudio \
  -H "Content-Type: application/json" \
  -d '{"finalVideo": "<final_video_path>", "songPath": "", "aiModel": "g4f", "musicSource": "library", "aspectRatio": "<ratio>"}'

Parameter References

Available Voices

Fetch from GET /api/models — response contains data.voices array.

Aspect Ratios

Supported Source Platforms

The backend uses yt-dlp as its download engine, which supports all of the above platforms natively.

Subtitle Templates

Defaults Summary