OFMOS — How to Create a Creator Concept

Core idea

Build a creator's content concept by training a fresh "research account" For You Page with lookalike creators, then mining their feeds for proven viral reels to recreate with the model. A video is only worth copying if it beat the account's own baseline (average of 6 videos before + 6 after) by at least 2x. On-screen text/SEO captions are essential to target the right audience (US/UK) instead of drifting to low-value geos.

Key concepts/frameworks

• FYP training: Use a dedicated research Instagram/TikTok account; like, comment, and share-to-self videos in the target niche until the FYP serves lookalike creators (~30–60 min).
• Lookalike matching: Match vibe/face/energy of the model ("not perfect, but similar"), not the exact niche — don't pick photographers just because they shot the style.
• 2x outlier rule: Baseline = avg views of the 6 videos before + 6 after the target; target must beat baseline ≥2x, else it's not an outlier. A 20x outlier is gold.
• On-screen text as SEO: Add a POV-style text headline so the algorithm targets the right audience; textless videos drift to India/low-value geos.
• Comment quality check: Verify comments are from target-market users (English, relatable: "gf take notes"); emoji/gif spam = bad reach despite views.
• Confusion-maxing ≠ good: High views + low comment count (e.g. 3M views/89 comments) = weak video; prefer 2M views/500+ comments.
• Trend freshness: Reels >3 weeks old risk being stale trends; prefer recent (under ~3 weeks).
• Keep it simple: Start with ONE creator's weekly plan before scaling to all models.

Action items

• Create a fresh research account; spend 30–60 min liking/commenting/sharing-to-self viral videos from lookalike creators to train the FYP.
• Put the model's account and candidate creators side by side; confirm vibe match before mining their content.
• For each candidate reel: check post date (skip >3 weeks), compute 6-before/6-after baseline, confirm ≥2x outlier, check comments for target-market quality.
• Check who lookalike creators follow to find more lookalikes.
• Plan adaptations: swap in model's assets (e.g. more cleavage, Latina angle), add US/World Cup jersey props, add transitions, add POV text headlines.
• Build a weekly content schedule for one creator first; find a week's worth of videos in a ~2-hour research session.
• Brief the model via short voice/video explanation → feed transcript to Claude → Claude outputs step-by-step instructions (makeup, lighting, framing, hook, background).
• Send briefs on a daily list or 2–3 day cadence; reduce direct calls to ~weekly.
• Accountability: give an underperforming new model 1 month of full effort, then drop and replace her.

SOPs/scripts/templates

• Quality gate doc (per-reel checklist): Does it match the creator's vibe? Did it overperform (≥2x baseline)? Is the hook good? Is it engaging? Can you add a viral factor? Can you do a comment bait? What to change for the recreation? Used as a training/recheck layer until the eye is internalized.
• Research session flow: train FYP → shortlist lookalikes → score reels against baseline → note "what to change" → compile into pattern bank / weekly schedule.

Tools

• "Shorts/Reels sort" browser extension (premium; works on TikTok + Instagram) — shows view counts, overperformance, and comment counts directly on profile grids.
• Instagram/TikTok — research accounts + FYP training; share-to-self feature used to train the algorithm fastest.
• Claude — turn voice-note briefs into step-by-step creator instructions.

Metrics/KPIs

• ≥2x over baseline = minimum to copy; 20x outlier = ideal.
• Comment health: ~500 comments/2M views good; 89 comments/3M views bad.
• Trend age: prefer <3 weeks old.
• Research time: FYP trainable in ~30 min; a week's content found in ~2 hours.