What viralari does.
Six things, all aimed at the same outcome: knowing why your best posts worked, and shipping more like them.
Hook intelligence
Every post's spoken hook gets transcribed and classified. Question, contrarian, listicle, story, vulnerable, authority, pattern interrupt - we name the type and score the strength. You see which hook types pull on your channel and which fall flat.
- Verbatim transcript of the first 10 seconds spoken on every reel
- On-screen text overlay extracted from the first 3 seconds
- Hook strength score 0-100 based on opening attention and re-watch potential
- Hook type classification across 10 categories
Pattern detection
We compare your top 10 winners against your bottom 10 underperformers and surface 5 to 8 channel-wide patterns. Posting times, content formats, hook types, caption structures, emotional registers, all named, ranked by effect size.
- Headline insight in plain English ('your talking-head giveaway videos posted Thursday evening run 6.4x baseline')
- Time-of-day heatmap from your real posting history
- Per-pattern confidence scores so you know which to trust
- Linked example posts so you can verify every claim
Voice fingerprint
We extract the phrases, openers, sentence rhythms, and rhetorical moves only you use. The fingerprint then seeds every hook generation so output reads like a good day of yours, not like generic AI.
- Common openers and closers pulled from your captions and transcripts
- Go-to rhetorical techniques (specificity, vulnerability, contrarian framing)
- Themes and topics you return to across the channel
- Per-channel voice profile, separate for IG and TikTok
Why it worked, why it didn't
Sonnet writes a per-post narrative for your top 10 winners and bottom 10 underperformers. Specific factors, not generic feedback. Linked side-by-side with classification, performance index, and lift vs baseline.
- Multi-paragraph 'why it worked' narrative on top winners
- Mirror narrative on underperformers, calling the specific failure modes
- Compare any two posts side-by-side with factor-by-factor diff
- Performance index that scores each post against its own peer cohort
Hook generation, in your voice
Tell us what you're filming. We generate 8 hook drafts using your winning patterns as the seed and your voice fingerprint as the style. Every output ships with a predicted score and reasoning.
- Spoken hook + suggested on-screen overlay per draft
- Predicted score and matched-pattern reasoning per hook
- Tone, length, and platform options
- History of every generation, with one-click rewrites
Honest about the limits
Instagram doesn't expose retention curves on reels or impressions on carousels. We tell you that, in plain English, instead of inventing fake retention scores. The numbers we show are real, the inferences are clearly flagged.
- Reels are scored on engagement per view, carousels on engagement per follower
- Each post is only compared against other posts of its own type
- Patterns flagged with confidence scores so you don't over-index on noise
- Caveats called out when the sample is too small for a claim to hold