Hooklayer
All tools/viral_remix3 creditssynthesize

What is viral_remix?

viral_remix is a Hooklayer MCP tool that takes a viral video (URL or transcript) and produces a fresh script mirroring its structural DNA — same scene count, same energy pattern, same beat-by-beat flow, different topic. Returns the extracted formula, scene-by-scene script with camera shots and text overlays, and a verify_hook block prompting independent score_hook verification.

Get free API keyTry in playground

No self-rating. viral_remix deliberately does not return a self-scored hook — that would be the script generator grading its own homework (cardinal coupling). Instead it returns verify_hook.hook_text + a recommendation to call score_hook for structurally-independent verification.

Phone-native filming only. Camera directions use "front-facing selfie", "phone propped on desk", "screen recording", "POV holding phone" — never cinematic terms like dolly, gimbal, rack focus. The output is shootable with just a phone.

Pass source_url OR transcript. URL extracts the transcript via Whisper automatically (adds ~3-5s latency). Transcript skips extraction. Cost: 3 credits regardless. Pairs with analyze_account.recommended_chain — the chain auto-fills viral_remix params with the top-replicability video URL.

Inputs & outputs

Endpoint: POST /api/v1/remix

Inputs

  • source_urlstring

    TikTok/YouTube/Instagram URL (one of source_url or transcript required)

  • transcriptstring

    Pre-extracted transcript — faster than URL extraction

  • my_topicstring

    What the remix should be about. Default: same niche as source

  • nichestring

    Optional niche hint

Output fields

  • extracted_formula

    hook_type, structure, pacing, emotional_triggers, why_it_works

  • fresh_script

    hook + scenes (section, voiceover, camera, action, text_overlay) + cta

  • camera_shots

    Flat list of camera directions (phone-native)

  • overlays

    Flat list of text overlays per scene

  • verify_hook

    hook_text + recommendation to call score_hook for independent verification

  • quality

    level + reason (degraded if <3 scenes or missing camera directions)

cURL

curl -X POST https://hooklayer.dev/api/v1/remix \
  -H "Authorization: Bearer hl_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "source_url": "https://www.tiktok.com/@user/video/1234567890",
    "my_topic": "B2B SaaS pricing tiers"
  }'

Example prompts

Paste any of these into Claude Desktop (with Hooklayer connected) to see the live response.

Remix a viral video for a new niche

Use Hooklayer viral_remix on this URL: https://www.tiktok.com/@humphreytalks/video/7188273048459857195 — remix the iPhone-shorting structure but for personal finance / index-fund education. Then call score_hook on the verify_hook.hook_text it returns to confirm the new hook scores ≥75.

Expected output: Returns 5-scene script keeping Humphrey's pattern-interrupt-then-reveal structure but rewriting the topic to compound investing. verify_hook prompts score_hook call; that scores ~80 because it inherits the proven structure.

Transcript-only remix (faster, no URL extraction)

I have the transcript of a viral video. Use Hooklayer viral_remix to remix it for the fitness niche. Transcript: "I'm gonna give you a thousand dollars. You can either spend it on an iPhone or put it in an index fund. One is worth $19,000 in 30 years..." Skip URL extraction — pass the transcript directly.

Expected output: Same structure, fitness topic. Skips Whisper transcription (~3-5s faster latency).

Chain after analyze_account

Run Hooklayer analyze_account on @humphreytalks. Take the source_url from his #2 video in recent_videos (sorted by engagement_score), then call viral_remix on it with my_topic="early-stage SaaS pricing decisions." Show me both outputs side by side.

Expected output: Demonstrates the recommended_chain pattern. analyze_account picks the highest-replicability video; viral_remix turns it into a SaaS-niche script with the same DNA.

Frequently asked

Why doesn't viral_remix return a hook score?

Cardinal coupling. When the same AI pipeline produces both the script and the score, the system collapses to a fixed point — every script lands at the model's "good but not great" mode. Hooklayer breaks the loop by deliberately NOT self-rating. Instead viral_remix returns a verify_hook block telling the agent to call score_hook for an independent assessment. Score from a structurally separate pipeline or don't score at all.

What does "viral DNA" mean here?

The structural beats that make a video work — hook type, scene count, pacing (HIGH→MEDIUM→RISING→PEAK→WARM), camera-switching cadence, emotional arc, payoff timing. These travel across topics. A 5-scene Mid-Action Drop with a reveal at scene 4 works whether the topic is investing or skincare. viral_remix extracts that skeleton and rebuilds it.

How is this different from ChatGPT writing me a TikTok script?

ChatGPT writes a script from a topic prompt. viral_remix writes a script from a proven-viral STRUCTURE — extracted from a real video with documented view counts. The output mirrors a viral skeleton, not a hypothetical template. Plus phone-native camera directions (no "dolly in" nonsense), text overlays per scene, and a verify_hook prompting independent scoring.

Can it handle videos longer than 60 seconds?

Yes — there's no hard cap. The output mirrors the structure of whatever source you pass. But for short-form (TikTok/Reels/Shorts) the model performs best on 15-90 second source videos; longer YouTube content extracts more pattern signal than you can fit in a 30-second remix.

What if the source URL fails to extract?

URL extraction uses Whisper for transcription. Failures (private videos, geo-blocked, removed) return a clear error with no credit deducted. Pass transcript directly as a fallback — copy the captions from the original video.

Does the remix copy the original copyrighted content?

No. viral_remix preserves the structural pattern (beat count, hook type, pacing) — which is not copyrightable — while completely rewriting the topic, dialogue, voiceover, and visuals. The output is original content built on a proven skeleton, the same way every cover song uses the same chord progressions.

Related tools

analyze_account
Deep TikTok creator analysis returning viral DNA scores, format fingerprint, steal map, and a recommended_chain for agent auto-firing.
score_hook
Score any TikTok hook 0-100 against viral patterns. Returns rewrites, calibration anchor, signals[] with evidence, and a counterfactual.
predict_virality
Score a draft script for viral potential with an adversarial check. Returns headline (skeptical) and optimistic scores, plus attack_vectors.

Try viral_remix in 30 seconds.

100 free credits at signup. No card. Works in Claude Desktop, Cursor, n8n, or any MCP client.

Get API keyInstall for Claude / Cursor / n8n