Problem

Small and mid-size YouTube/TikTok creators (1,000–100,000 subscribers) lose income because sponsor pitching is guesswork: they pitch brands whose audiences don't match, then get rejected or paid poorly. Small brands also have no data to judge which creators' audiences truly like them. Nobody has match evidence — only follower counts and gut feeling. PitchTaste provides that evidence.

What was built

A web app. The creator picks a ready-made archetype (true-crime documentary channel, cozy game streamer, beginner finance TikToker) or types 3 creators/shows they admire plus their niche, then clicks "Cari sponsor" (Find sponsors). Result: a ranked brand shortlist, an evidence panel per brand, and a ready-to-copy pitch draft per brand.

How Qloo is used

This is the core, precisely:

  1. /search turns creator/show names into valid Qloo entity IDs.
  2. /v2/tags turns the niche into valid Qloo tag IDs (no invented tags are passed through).
  3. /v2/insights with filter.type=brand, signal.interests.entities = the creator's taste seeds, feature.explainability=true — returns the ranked brand shortlist plus per-seed contribution metadata (scores 0–1).
  4. NO LLM decides the match. The pitch generator only fills slots from Qloo evidence (brand, niche, top seed, affinity score). No evidence = no pitch.

Why Qloo is irreplaceable

Without Qloo's taste graph, this product collapses completely — an LLM could only guess "this brand feels like a fit" with no audience-affinity data at all. The only match evidence (fans of X also over-index on brand Y, with explainability scores) exists only in Qloo. Turn Qloo off, the shortlist disappears.

Request to result (sanitized, no credentials)

Input (taste seeds + niche) → entity/tag ID resolution → one insights call (brand filter + seed signal + explainability) → ranking + evidence panel → slot-filled pitch. All Qloo calls go through a server-side proxy; the API key is never in the repo, client bundle, or demo recording (server env var only). No personal data is sent to Qloo — only public cultural taste names. Results are aggregate affinity ("fans of [seed] also over-index on [brand]"), not individual profiles.

Known limitations

Indonesian local brand coverage in Qloo can be thin; the default demo uses global brands + global archetypes, clearly labeled. If Qloo returns empty, the app honestly says "no strong match" instead of inventing one. Qloo results are aggregate cultural affinity, not proof of identity, preference, or future behavior of any person. The pitch generator deliberately has no LLM deciding matches (evidence-slot templates only).Built with: Next.js, Qloo Taste AI API, Vercel.

Built With

  • next.js
  • qloo
  • vercel
Share this project:

Updates

Submission history