How it works
How Hookquill learns your channel’s voice
Hookquill learns your voice from your own videos. Paste your channel link and it reads the captions of up to 20 long-form videos: your most-watched and your newest. It measures your sentence length and speaking pace, describes how you talk, and gives every script it writes about 1,300 words of your real speech to match.
By Founder @ Hookquill · Updated
What it reads
- Your latest 30 long-form uploads. Shorts are skipped.
- The captions of up to 20 of them: your 15 most-watched and your 5 newest. Videos without captions are skipped, and you see how many it could read.
- Reading a channel is free. Building the voice profile from it costs 20 credits.
- No channel yet? It learns from a short questionnaire and anything you have written: a post, a newsletter, captions, a script you filmed.
What it learns
- How you talk, in plain words: the role you play for viewers, your tone, your reading level and how you address them.
- The phrases you really repeat. A phrase is kept only if it appears in your own material.
- Your average sentence length and your speaking pace, counted from your captions rather than guessed.
- How your videos are built: the way you open, the order of your sections, your pacing.
- About 1,300 words of your own speech, word for word: the openings of your three most-watched videos and a passage from the middle of the top two. Punctuation is restored only where captions had none, and a passage keeps its raw text if a single word would change.
How every script uses it
Every writing step follows one order. Your channel rules come first: an always and never box you fill in, such as "never mention competitors". Then your real passages, then the profile, then Hookquill’s own defaults, which only fill gaps. Only the rules about facts and safety outrank you.
When you reject a style edit (More like me, Punchier, Simpler, Change tone), later scripts take note. Factual corrections never change your voice.
Your channel, never someone else’s
Channels you admire teach structure only: how their videos open, how sections are ordered, how fast they move. Their words never reach your voice profile; that rule is enforced in the code, not left to the model. Hookquill does not imitate other creators and does not clone voices.
How sure it is
- Strong match: ten or more of your own videos were read.
- Good match: some of your videos, or at least two pieces of your own writing.
- Still learning: a questionnaire and channels you like, with nothing of your own yet.
What it will not do
It does not invent your life. No first-person stories, results or opinions you have not given it; where a personal story would help, it uses a sourced example or a clearly framed hypothetical. After the draft, every AI change arrives as a suggestion you accept or reject line by line.
Captions are YouTube data: kept for up to 30 days, then refreshed or deleted. Your scripts are yours, and they are never used to train AI models.
Questions
How many videos does Hookquill need to learn my voice?
Ten captioned videos give a strong match. A few videos, or two or three pieces of your own writing, give a good one. A brand-new channel can start from a short questionnaire and sharpen the profile later.
Can it learn from a faceless channel with an AI or hired narrator?
Yes. The captions of a faceless video are the script the narrator read, which is exactly what Hookquill learns from. Videos without captions are skipped.
Can I make my scripts sound like another YouTuber?
No. Channels you like teach structure only. Their words never feed your voice, and Hookquill does not imitate other creators.
Can each of my channels have its own voice?
Yes. Each channel gets its own voice profile, learned from that channel’s videos: 3 on Starter, 10 on Creator, unlimited on Pro.
Something here wrong or unclear? Email us and the page gets fixed.