Grade your faceless video hook before you publish.

The score (0–100) estimates how likely your opener is to grab and hold attention in the first 3–10 seconds. It's based on a fixed 5-factor rubric: clarity, curiosity gap, specificity, payoff strength, and early-hold outlook. For a deeper walkthrough see how to write a Shorts script that hooks in 3 seconds.

It's a diagnostic, not a prediction. The rubric is derived from the retention patterns we mapped in high-performing short-form openers, so it'll catch structural weaknesses reliably. But no tool can guarantee performance — treat it as a second opinion, not a verdict.

Any short-form video opener: YouTube Shorts, TikTok, Instagram Reels, long-form YouTube intros. Designed for faceless formats but works fine for any creator.

YouTube publishes no 7 second rule for hooks. The seven seconds in its advertiser-friendly guidelines apply only to graphic game violence at the start of a video. What YouTube does show is "How many chose to view" in YouTube Studio: how often viewers watched your Short instead of swiping away. Compare it across your own Shorts. More in how to write a Shorts script that hooks in 3 seconds.

Ten per day, free. No signup, no card. The limit exists to keep the tool fast for everyone. Need more? The full ViralFaceless app turns a script into a finished 9:16 video; see where an AI video factory breaks for the upgrade rationale.

Hooks are sent to the grading service to be scored and aren't saved to any account or shared. We keep standard request logs for reliability and abuse prevention, but nothing is used to train models or build a hook corpus.

ChatGPT will give you plausible-sounding feedback, but it doesn't use a fixed rubric, doesn't benchmark against retention patterns, and will give you different answers on different days. This is built around a consistent 5-factor diagnostic.

Alongside the 5-factor score, the grader detects a set of core attention triggers directly in your hook text — deterministically, not by vibes. It checks for an open loop, an exact number, an audience marker, a pattern break, a mechanism reveal, a competence check, and a second-person moment. Fired triggers show as a check; missing ones come with a one-line note on how to add them. Title mode uses the applicable subset — the triggers that earn a click on a scanned feed (open loop, exact number, audience marker, pattern break, and competence check).

Switch the toggle to Title and the grader judges click-through instead of in-video retention. It scores the same five factors but reweights them for how a title earns the click on a scanned feed or search page — curiosity gap and specificity carry more, and the fifth factor is relabeled "Scan-Stop Impression". Rewrites come back as titles (short, no clickbait), and the trigger checklist narrows to the ones that move title click-through. Title mode works in English and Russian — see the languages question below.

Hooks are graded in English. Title mode works in English and Russian — the deterministic signals (numbers, negative framing, questions, comparisons, niche markers, clickbait tells) fire on Russian titles too, and the written feedback and rewrites come back in the same language as the title you paste. Other languages will still score on the language-neutral signals (digits, question marks, length), but the written feedback is tuned for English and Russian.

The default title scale is calibrated for feed-scan Shorts conventions, where a title over ~65 characters gets penalized as too long. The story-long profile switches to the long-story-video scale, where the title is thumbnail text and 90–100 characters is the normal, optimal length — a long title is no longer penalized. It also counts the «X — а Y…» / «пока не…» reversal tease (a premise that flips after a completed clause, or a trailing «…») as a real curiosity signal, and treats a mid-title CAPS burst as emphasis rather than shouting. Same five factors and the same 0–100 scale — a good niche story title honestly lands in the Strong 60–80 range, while clickbait still scores low. The scale meaning is per-profile: a 70 on the default scale and a 70 under story-long are read against different length and curiosity conventions.