ahmedsakri/youtube-agent-pro / openai/skills/yt-voice
yt-voice
VerifiedBuild or refresh an optional creator voice profile from their own transcripts and stated audience, vocabulary, and format preferences. Use for "set up my voice", "write my voice.md", "make the scripts sound like me", "my scripts don't sound like me", or first-time setup of the YouTube pack.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 3.0 KB |
| agents/openai.yaml | config | 240 B |
| assets/voice.md | markdown | 1.1 KB |
| voiceprint.py | script | 5.0 KB |
voiceprint.py
View on GitHub#!/usr/bin/env python3
"""voiceprint.py - read your own transcripts and measure how you actually talk.
python3 voiceprint.py best-video.txt
python3 voiceprint.py vid1.srt vid2.srt vid3.txt # merge several, the more the better
python3 voiceprint.py *.txt --json
Point it at the transcripts of your best videos (not your scripts - the words you actually said) and
it measures the things a voice.md should capture but that you cannot see about yourself: your mean
sentence length, your filler rate, the distinctive words you lean on, and the hype words to flag.
It does not write your voice.md - it gives yt-voice the measurements to write it from.
SENTENCE LENGTH short and punchy, or long and winding. A script that fights this is unreadable
out loud.
FILLER RATE how much "like / basically / literally / you know" you run - fine to keep some,
useful to know the baseline.
SIGNATURE WORDS content words you use far more than normal speech would. These are your tics.
Keep them; they are what makes a script sound like you and not like a blog.
HYPE WORDS "insane / crazy / game-changer" you already say - decide which stay and which
go in the banned list.
IT MEASURES, IT DOES NOT JUDGE. A long mean sentence is not worse than a short one. A high filler
rate is not a grade. This is a mirror, so the script a skill writes you is readable on the first
take instead of the fourth.
"""
import argparse, json, os, re, sys
from collections import Counter
STOP = set("""a an and are as at be been but by for from had has have he her his i if in is it its
of on or our she that the their them they this to was we were what when which who will with you your
yours me my mine us do does did not no yes so just can could would should now then there here all
about into over after before out up down off than too very can't dont don't im i'm it's thats
that's youre you're were we're they're he's she's""".split())
FILLER = {"like", "basically", "literally", "actually", "really", "just", "you know", "i mean",
"sort of", "kind of", "um", "uh", "right", "okay", "so", "well", "stuff", "things"}
HYPE = {"insane", "crazy", "amazing", "incredible", "huge", "massive", "game-changer", "gamechanger",
"mindblowing", "unbelievable", "epic", "ultimate", "revolutionary", "perfect", "literally"}
TS = re.compile(r"^\s*\d+\s*$|-->|\d{1,2}:\d{2}")
def read(path):
lines = []
for raw in open(path, encoding="utf-8", errors="replace").read().splitlines():
if TS.search(raw) or not raw.strip():
continue
lines.append(raw.strip())
return " ".join(lines)
def analyse(text):
low = text.lower()
tokens = re.findall(r"[a-z']+", low)
n = len(tokens)
sentences = [s for s in re.split(r"[.!?]+", text) if s.strip()]
slens = [len(re.findall(r"[a-z']+", s.lower())) for s in sentences] or [0]
mean_sentence = sum(slens) / len(slens)
filler_hits = 0
for f in FILLER:
filler_hits += len(re.findall(r"\b" + re.escape(f) + r"\b", low)) if " " in f else tokens.count(f)
filler_rate = filler_hits / n * 100 if n else 0
content = [t for t in tokens if t not in STOP and len(t) > 3]
freq = Counter(content)
# signature = frequent content words, normalised by how long they are (longer = more distinctive)
signature = [w for w, _ in sorted(freq.items(), key=lambda kv: -(kv[1] * (1 + len(kv[0]) / 10)))[:15]]
hype_present = sorted({t for t in tokens if t in HYPE})
return {
"words": n,
"sentences": len(sentences),
"mean_sentence_words": round(mean_sentence, 1),
"filler_rate_pct": round(filler_rate, 2),
"signature_words": signature,
"hype_words_you_already_use": hype_present,
}
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("transcripts", nargs="+")
parser.add_argument("--json", action="store_true")
args = parser.parse_args()
files, as_json = args.transcripts, args.json
try:
text = " ".join(read(f) for f in files)
except OSError as exc:
parser.error(str(exc))
if not text.strip():
parser.error("transcripts contain no spoken text")
r = analyse(text)
if as_json:
print(json.dumps(r, indent=1))
return
print(f"\n {len(files)} transcript(s) {r['words']:,} words {r['sentences']} sentences\n")
print(f" mean sentence {r['mean_sentence_words']} words "
+ ("short, punchy" if r['mean_sentence_words'] < 12 else
"long, winding - write to it, do not fight it" if r['mean_sentence_words'] > 20 else "mid"))
print(f" filler rate {r['filler_rate_pct']}% of all words")
print(f" signature words {', '.join(r['signature_words'][:12])}")
print(f" hype you use {', '.join(r['hype_words_you_already_use']) or '(none - clean)'}")
print("\n hand these to yt-voice to write your chosen profile file in your actual voice.\n")
if __name__ == "__main__":
main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |