Menu
Expert knowledge for AI video clipping โ yt-dlp downloading, whisper transcription, SRT generation, and ffmpeg processing
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | clip-hand-skill |
| version | 2.0.0 |
| description | Expert knowledge for AI video clipping โ yt-dlp downloading, whisper transcription, SRT generation, and ffmpeg processing |
| runtime | prompt_only |
All tools (ffmpeg, ffprobe, yt-dlp, whisper) use identical CLI flags on Windows, macOS, and Linux. The differences are only in shell syntax:
| Feature | macOS / Linux | Windows (cmd.exe) |
|---|---|---|
| Suppress stderr | 2>/dev/null | 2>NUL |
| Filter output | | grep pattern | | findstr pattern |
| Delete files | rm file1 file2 | del file1 file2 |
| Null output device | -f null - | -f null - (same) |
| ffmpeg subtitle paths | subtitles=clip.srt | subtitles=clip.srt (relative OK, absolute needs C\\:/path) |
IMPORTANT: ffmpeg filter paths (-vf "subtitles=...") always need forward slashes. On Windows with absolute paths, escape the colon: subtitles=C\\:/Users/me/clip.srt
Prefer using file_write tool for creating SRT/text files instead of shell echo/heredoc.
# Best video up to 1080p + best audio, merged
yt-dlp -f "bv[height<=1080]+ba/b[height<=1080]" --restrict-filenames -o "source.%(ext)s" "URL"
# 720p max (smaller, faster)
yt-dlp -f "bv[height<=720]+ba/b[height<=720]" --restrict-filenames -o "source.%(ext)s" "URL"
# Audio only (for transcription-only workflows)
yt-dlp -x --audio-format wav --restrict-filenames -o "audio.%(ext)s" "URL"
# Get full metadata as JSON (duration, title, chapters, available subs)
yt-dlp --dump-json "URL"
# Key fields: duration, title, description, chapters, subtitles, automatic_captions
# Download auto-generated subtitles in json3 format (word-level timing)
yt-dlp --write-auto-subs --sub-lang en --sub-format json3 --skip-download --restrict-filenames -o "source" "URL"
# Download manual subtitles if available
yt-dlp --write-subs --sub-lang en --sub-format srt --skip-download --restrict-filenames -o "source" "URL"
# List available subtitle languages
yt-dlp --list-subs "URL"
--restrict-filenames โ safe ASCII filenames (no spaces/special chars) โ important on all platforms--no-playlist โ download single video even if URL is in a playlist-o "template.%(ext)s" โ output template (%(ext)s auto-detects format)--cookies-from-browser chrome โ use browser cookies for age-restricted content--extract-audio / -x โ extract audio only--audio-format wav โ convert audio to wav (for whisper)# Extract mono 16kHz WAV (whisper's preferred input format)
ffmpeg -i source.mp4 -vn -ar 16000 -ac 1 -y audio.wav
# Standard transcription with word-level timestamps
whisper audio.wav --model small --output_format json --word_timestamps true --language en
# Faster alternative (same flags, 4x speed)
whisper-ctranslate2 audio.wav --model small --output_format json --word_timestamps true --language en
| Model | VRAM | Speed | Quality | Use When |
|---|---|---|---|---|
| tiny | ~1GB | Fastest | Rough | Quick previews, testing pipeline |
| base | ~1GB | Fast | OK | Short clips, clear speech |
| small | ~2GB | Good | Good | Default โ best balance |
| medium | ~5GB | Slow | Better | Important content, accented speech |
| large-v3 | ~10GB | Slowest | Best | Final production, multiple languages |
Note: On macOS Apple Silicon, consider mlx-whisper as a faster native alternative.
{
"text": "full transcript text...",
"segments": [
{
"id": 0,
"start": 0.0,
"end": 4.52,
"text": " Hello everyone, welcome back.",
"words": [
{"word": " Hello", "start": 0.0, "end": 0.32, "probability": 0.95},
{"word": " everyone,", "start": 0.32, "end": 0.78, "probability": 0.91},
{"word": " welcome", "start": 0.78, "end": 1.14, "probability": 0.98},
{"word": " back.", "start": 1.14, "end": 1.52, "probability": 0.97}
]
}
]
}
segments[].words[] gives word-level timing when --word_timestamps trueprobability indicates confidence (< 0.5 = likely wrong){
"events": [
{
"tStartMs": 1230,
"dDurationMs": 5000,
"segs": [
{"utf8": "hello ", "tOffsetMs": 0},
{"utf8": "world ", "tOffsetMs": 200},
{"utf8": "how ", "tOffsetMs": 450},
{"utf8": "are you", "tOffsetMs": 700}
]
}
]
}
For each event and each segment within it:
word_start_ms = event.tStartMs + seg.tOffsetMsword_start_secs = word_start_ms / 1000.0word_text = seg.utf8.trim()Events without segs are line breaks or formatting โ skip them.
Events with segs containing only "\n" are newlines โ skip them.
1
00:00:00,000 --> 00:00:02,500
First line of caption text
2
00:00:02,500 --> 00:00:05,100
Second line of caption text
HH:MM:SS,mmm (comma separator, not dot)file_write tool to create the SRT file โ works identically on all platformsFor animated/styled captions, use ASS subtitle format instead of SRT:
ffmpeg -i clip.mp4 -vf "subtitles=clip.ass:force_style='FontSize=22,FontName=Arial,Bold=1,PrimaryColour=&H00FFFFFF,OutlineColour=&H00000000,Outline=2,Shadow=1,Alignment=2,MarginV=40'" -c:a copy output.mp4
Key ASS style properties:
PrimaryColour=&H00FFFFFF โ white text (AABBGGRR format)OutlineColour=&H00000000 โ black outlineOutline=2 โ outline thicknessAlignment=2 โ bottom centerMarginV=40 โ margin from bottom edgeFontSize=22 โ good size for 1080x1920 verticalffmpeg -i input.mp4 -filter:v "select='gt(scene,0.3)',showinfo" -f null - 2>&1
pts_time: from showinfo output for timestampsgrep showinfo, on Windows pipe through findstr showinfoffmpeg -i input.mp4 -af "silencedetect=noise=-30dB:d=1.5" -f null - 2>&1
d=1.5 = minimum 1.5 seconds of silencesilence_start and silence_end in output# Re-encoded (accurate cuts)
ffmpeg -ss 00:01:30 -to 00:02:15 -i input.mp4 -c:v libx264 -c:a aac -preset fast -crf 23 -movflags +faststart -y clip.mp4
# Lossless copy (fast but may have keyframe alignment issues)
ffmpeg -ss 00:01:30 -to 00:02:15 -i input.mp4 -c copy -y clip.mp4
-ss before -i = fast seek (recommended for extraction)-to = end timestamp, -t = duration# Center crop (when source is 16:9)
ffmpeg -i input.mp4 -vf "crop=ih*9/16:ih:(iw-ih*9/16)/2:0,scale=1080:1920" -c:a copy output.mp4
# Scale with letterbox padding (preserves full frame)
ffmpeg -i input.mp4 -vf "scale=1080:1920:force_original_aspect_ratio=decrease,pad=1080:1920:(ow-iw)/2:(oh-ih)/2:black" -c:a copy output.mp4
# SRT subtitles with styling (use relative path or forward-slash absolute path)
ffmpeg -i input.mp4 -vf "subtitles=subs.srt:force_style='FontSize=22,FontName=Arial,PrimaryColour=&H00FFFFFF,OutlineColour=&H00000000,Outline=2,Alignment=2,MarginV=40'" -c:a copy output.mp4
# Simple text overlay
ffmpeg -i input.mp4 -vf "drawtext=text='Caption':fontsize=48:fontcolor=white:borderw=3:bordercolor=black:x=(w-text_w)/2:y=h-th-40" output.mp4
Windows path escaping: subtitles=C\\:/Users/me/subs.srt (double-backslash before colon)
# At specific time (2 seconds in)
ffmpeg -i input.mp4 -ss 2 -frames:v 1 -q:v 2 -y thumb.jpg
# Best keyframe
ffmpeg -i input.mp4 -vf "select='eq(pict_type,I)',scale=1280:720" -frames:v 1 thumb.jpg
# Contact sheet
ffmpeg -i input.mp4 -vf "fps=1/10,scale=320:-1,tile=4x4" contact.jpg
# Full metadata (JSON)
ffprobe -v quiet -print_format json -show_format -show_streams input.mp4
# Duration only
ffprobe -v error -show_entries format=duration -of csv=p=0 input.mp4
# Resolution
ffprobe -v error -select_streams v:0 -show_entries stream=width,height -of csv=p=0 input.mp4
Fastest cloud STT โ uses whisper-large-v3 on Groq hardware. Free tier available.
curl -s -X POST "https://api.groq.com/openai/v1/audio/transcriptions" \
-H "Authorization: Bearer $GROQ_API_KEY" \
-H "Content-Type: multipart/form-data" \
-F "file=@audio.wav" \
-F "model=whisper-large-v3" \
-F "response_format=verbose_json" \
-F "timestamp_granularities[]=word" \
-o transcript_raw.json
Response: {"text": "...", "words": [{"word": "hello", "start": 0.0, "end": 0.32}]}
timestamp_granularities[]=word is required for word-level timing.curl -s -X POST "https://api.openai.com/v1/audio/transcriptions" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: multipart/form-data" \
-F "file=@audio.wav" \
-F "model=whisper-1" \
-F "response_format=verbose_json" \
-F "timestamp_granularities[]=word" \
-o transcript_raw.json
Response format same as Groq. Max 25MB.
curl -s -X POST "https://api.deepgram.com/v1/listen?model=nova-2&smart_format=true&utterances=true&punctuate=true" \
-H "Authorization: Token $DEEPGRAM_API_KEY" \
-H "Content-Type: audio/wav" \
--data-binary @audio.wav \
-o transcript_raw.json
Response: {"results": {"channels": [{"alternatives": [{"words": [{"word": "hello", "start": 0.0, "end": 0.32, "confidence": 0.99}]}]}]}}
smart_format=true adds punctuation and casing.# List available voices
edge-tts --list-voices
# Generate speech
edge-tts --text "Your caption text here" --voice en-US-AriaNeural --write-media tts_output.mp3
# Other good voices: en-US-GuyNeural, en-GB-SoniaNeural, en-AU-NatashaNeural
Install: pip install edge-tts
curl -s -X POST "https://api.openai.com/v1/audio/speech" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"tts-1","input":"Your text here","voice":"alloy"}' \
--output tts_output.mp3
Voices: alloy, echo, fable, onyx, nova, shimmer
Models: tts-1 (fast), tts-1-hd (quality)
curl -s -X POST "https://api.elevenlabs.io/v1/text-to-speech/21m00Tcm4TlvDq8ikWAM" \
-H "xi-api-key: $ELEVENLABS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text":"Your text here","model_id":"eleven_monolingual_v1"}' \
--output tts_output.mp3
Voice ID 21m00Tcm4TlvDq8ikWAM = Rachel (default). List voices: GET /v1/voices
# Mix TTS over original audio (original at 30% volume, TTS at 100%)
ffmpeg -i clip.mp4 -i tts.mp3 \
-filter_complex "[0:a]volume=0.3[orig];[1:a]volume=1.0[tts];[orig][tts]amix=inputs=2:duration=first[out]" \
-map 0:v -map "[out]" -c:v copy -c:a aac -y clip_voiced.mp4
# Replace audio entirely (no original audio)
ffmpeg -i clip.mp4 -i tts.mp3 -map 0:v -map 1:a -c:v copy -c:a aac -shortest -y clip_voiced.mp4
-preset ultrafast for quick previews, -preset slow for final output-crf 23 for good quality (18=high, 28=low, lower=bigger files)-movflags +faststart for web-friendly MP4-threads 0 to auto-detect CPU cores-y to overwrite without askingcurl -s -X POST "https://api.telegram.org/bot<BOT_TOKEN>/sendVideo" \
-F "chat_id=<CHAT_ID>" \
-F "video=@clip_N_final.mp4" \
-F "caption=Clip title here" \
-F "parse_mode=HTML" \
-F "supports_streaming=true"
| Parameter | Required | Description |
|---|---|---|
chat_id | Yes | Channel (-100XXXXXXXXXX or @channelname), group, or user numeric ID |
video | Yes | @filepath for upload (max 50MB) or a Telegram file_id for re-send |
caption | No | Text caption, up to 1024 characters |
parse_mode | No | HTML or MarkdownV2 for styled captions |
supports_streaming | No | true enables progressive playback |
{"ok": true, "result": {"message_id": 1234, "video": {"file_id": "BAACAgI...", "file_size": 5242880}}}
{"ok": false, "error_code": 400, "description": "Bad Request: chat not found"}
| Error Code | Description | Fix |
|---|---|---|
| 400 | Chat not found | Verify chat_id; bot must be added to the channel/group |
| 401 | Unauthorized | Bot token is invalid or revoked โ regenerate via @BotFather |
| 413 | Request entity too large | File exceeds 50MB โ re-encode: ffmpeg -i input.mp4 -fs 49M -c:v libx264 -crf 28 -preset fast -c:a aac -y output.mp4 |
| 429 | Too many requests | Rate limited โ wait the retry_after seconds from the response |
Telegram allows up to 50MB for video uploads via Bot API. If a clip exceeds this:
ffmpeg -i clip_N_final.mp4 -fs 49M -c:v libx264 -crf 28 -preset fast -c:a aac -movflags +faststart -y clip_N_tg.mp4
WhatsApp Cloud API requires uploading the video first to get a media_id, then sending a message referencing that ID.
curl -s -X POST "https://graph.facebook.com/v21.0/<PHONE_NUMBER_ID>/media" \
-H "Authorization: Bearer <ACCESS_TOKEN>" \
-F "file=@clip_N_final.mp4" \
-F "type=video/mp4" \
-F "messaging_product=whatsapp"
Success response:
{"id": "1234567890"}
curl -s -X POST "https://graph.facebook.com/v21.0/<PHONE_NUMBER_ID>/messages" \
-H "Authorization: Bearer <ACCESS_TOKEN>" \
-H "Content-Type: application/json" \
-d '{
"messaging_product": "whatsapp",
"to": "<RECIPIENT_PHONE>",
"type": "video",
"video": {
"id": "<MEDIA_ID>",
"caption": "Clip title here"
}
}'
Success response:
{"messaging_product": "whatsapp", "contacts": [{"wa_id": "14155551234"}], "messages": [{"id": "wamid.HBgL..."}]}
WhatsApp allows up to 16MB for video uploads. If a clip exceeds this:
ffmpeg -i clip_N_final.mp4 -fs 15M -c:v libx264 -crf 30 -preset fast -c:a aac -movflags +faststart -y clip_N_wa.mp4
WhatsApp requires the recipient to have messaged you within the last 24 hours (for non-template messages). If you get a "template required" error, either:
| Error Code | Description | Fix |
|---|---|---|
| 100 | Invalid parameter | Check phone_number_id and recipient format (no + prefix, no spaces) |
| 190 | Invalid/expired access token | Regenerate token in Meta Business Settings; temporary tokens expire in 24h |
| 131030 | Recipient not in allowed list | In test mode, add recipient to allowed numbers in Meta Developer Portal |
| 131047 | Re-engagement message / template required | Recipient hasn't messaged within 24h โ use a template or ask them to message first |
| 131053 | Media upload failed | File too large or unsupported format โ re-encode as MP4 under 16MB |
Privacy-respecting metasearch specialist using SearXNG instances
Playwright-based browser automation patterns for autonomous web interaction
Expert knowledge for the Infisical Sync Hand โ Infisical API reference, vault operations, error patterns, security guidance
Expert knowledge for AI deep research โ methodology, source evaluation, search optimization, cross-referencing, synthesis, and citation formats
Expert knowledge for autonomous market intelligence and trading โ technical analysis, risk management, Alpaca API, financial data sources
Expert knowledge for AI Twitter/X management โ API v2 reference, content strategy, engagement playbook, safety, and performance tracking