Our methodology

Every video in Viral Anatomy goes through VidMojo's automated analysis pipeline:

1. 1
   Frame extraction — We sample one keyframe per second from the source video. These still images — never the original video file — are what you see in the shot grid.
2. 2
   Shot-level AI vision — Each frame is passed through an AI vision model (GPT-4o). The model describes subjects, setting, mood, on-screen text, and visual style. It never has access to user data or anything beyond the frame image.
3. 3
   Transcription — Audio is transcribed with OpenAI Whisper. We store only the text — we do not retain the original audio.
4. 4
   Viral signal synthesis — A second AI pass reads the full frame sequence and transcript together to identify the specific moments that likely drove engagement: the hook structure, the pattern interrupt, the payoff. These become the viral signals you see on each page.
5. 5
   Recreation prompt generation — A final pass produces an AI-ready prompt describing the video's format, pacing, and structure — so creators can make their own version. The original creator's script or exact words are never reproduced verbatim.

Every Viral Anatomy page displays the creator's TikTok handle and links directly to the original video on TikTok. We believe analysis is only useful in context of the creator who made it.

• Creator handle and TikTok profile link appear prominently at the top of every page.
• We link to the original TikTok video, not a re-hosted copy.
• We do not host, stream, or embed the original video. No iframes are used.
• All frame images are still keyframes, not playable video content.

Creators own their content. If you want your video removed from Viral Anatomy — for any reason — we will honor that request within 48 hours.

To request removal, use the “Request takedown” link at the bottom of the relevant page, or email [email protected].

We do not require creators to provide legal justification. A simple “please remove my video” is sufficient.

Viral Anatomy is a commentary and educational platform. Our analysis is transformative — it adds frame-level annotation, structural observation, and educational context that isn't present in the original work.

We rely on fair use principles where applicable. We do not reproduce creative expression verbatim; we analyze structure and pattern. Recreation prompts are generalized format descriptions, not transcripts.

If you are a creator or rights holder with a concern, please contact us. We take all concerns seriously and respond within 48 hours.

All analysis is AI-assisted. AI models can misidentify subjects, misread on-screen text, or attribute viral success to elements that weren't actually causal. We display an inline transparency card on every breakdown page.

If you believe an analysis contains factually inaccurate information, you can report it using the takedown form on the relevant page. We review reported inaccuracies and update or remove content within 48 hours.