entit.ai

// x402 · pay per analysis · no api keys

Structured
intelligence
from any
short-form video.

POST a TikTok, Reels, or Shorts URL. Pay one micropayment. Get default VideoAnalysis JSON — or pass format for one of 24 typed workflow templates: recipe, creator brief, content screening, and more.

// SUPPORTED PLATFORMS
TIKTOKINSTAGRAM REELSYOUTUBE SHORTS

// §02 · DEFAULT OUTPUT · VideoAnalysis

Audio + visual
understanding.

Not transcription-only. When you POST a URL without a format, every analysis combines the six fields below in one structured JSON response. Want a different shape? See workflow formats below.

01transcription

Speech transcription

Full spoken-word transcript from the video audio track.

02language

Language detection

ISO language code for the primary spoken language.

03onScreenText

On-screen text

OCR-style extraction of text overlays and captions.

04visualDescription

Visual summary

Narrative description of what happens on screen.

05scenes[]

Scene timeline

Timestamped scene breakdowns for navigation and search.

06author / authorHandle

Creator metadata

Author name, handle, title, and platform when available.

// §03 · WORKFLOW FORMATS

One API.
24 output shapes.

Pass optional format on job submit to get typed JSON for a specific workflow — recipes, creator briefs, workout plans, shop-the-look, and more. Same price as default analysis; caches are isolated per format.

Brandcontent-screening

Brand safety, UGC compliance, ad-placement suitability

Brandcreator-brief

Hooks, angle, key moments, influencer vetting

Foodrecipe

Ingredients and steps from cooking clips

Contenttutorial-guide

How-to videos → written guide / FAQ

Contentquote-highlights

Shareable quotes and highlight moments

E-commerceproduct-tagging

Products shown with search queries

E-commerceshop-the-look

Fashion items and outfit search queries

E-commercereview-summary

Pros/cons from product review videos

Healthworkout-plan

Exercises with sets, reps, rest

Healthhealth-claims

Medical claims flagged for fact-checking

Educationflashcards

Study flashcards from educational content

Educationcurriculum-tags

Subject, difficulty, learning objectives

Educationlanguage-learning

Phrases, vocabulary, grammar patterns

Newsugc-verification

Scene/location hints for citizen journalism checks

Newsbreaking-news-signal

Newsworthiness signal from one clip

Newseyewitness-timeline

Structured timeline of factual claims

Musicsong-context

Background music / performance for licensing

Musiclyric-hook

Lip-sync, hooks, rights notes

Musicsound-trend

Audio description and viral potential

Sportstechnique-analysis

Coaching cues for technique clips

Sportshighlight-reel

Action-dense clip timestamps

Sportsscouting-report

Athlete performance notes for coaches

Traveldestination-guide

Itinerary, tips, costs from travel vlogs

Travelrestaurant-discovery

Dishes, prices, atmosphere from food reviews

// AGENT DISCOVERY · NATURAL LANGUAGE → FORMAT

Agents and search tools can map user intent to a format slug. Examples:

  • convert a cooking YouTube Short or TikTok to a text recipe→ format: recipe
  • turn a how-to Reel into written steps or documentation→ format: tutorial-guide
  • extract pull quotes from a short-form video for social posts→ format: quote-highlights
  • brand safety check before reposting UGC→ format: content-screening
  • influencer vetting brief with hooks and key moments→ format: creator-brief
  • find products shown in a shopping haul video→ format: product-tagging
  • shop the outfit from a fashion Reel→ format: shop-the-look
  • summarize pros and cons from a product review Short→ format: review-summary

Full intent list in llms.txt and the agent integration guide.

// LIMITATIONS · WHAT FORMATS DO NOT DO

  • Single video per call

    Each job analyses one short-form video. Feed-scale monitoring across many clips is achieved by orchestrating many jobs.

  • E-commerce returns queries, not links

    product-tagging, shop-the-look, and review-summary return searchQuery strings — not buyable URLs or affiliate links.

  • Claims are flagged, not verified

    Health, news, and ugc-verification formats surface claims for human review. They do not assert that any claim is true.

  • Every format has a published schema

    All 24 formats have JSON Schemas on /docs/response-schema, auto-generated from the Gemini contract in internal/gemini/formats. 3 (content-screening, creator-brief, recipe) also ship with hand-curated Go structs and field tables; the rest are typed by JSON Schema only. OpenAPI still references WorkflowFormatResult as a fallback while typed schemas are promoted into openapi.json.

// §04 · INTEGRATION

How it works.

Async job model. Pay once on submit; poll until status: success.

01

Submit URL

PAID · x402

POST /v1/jobs with a video URL and an x402 payment header. Returns a jobId.

02

Poll job

FREE

GET /v1/jobs/{jobId} is free. Server returns Retry-After: 5 while processing.

03

Receive JSON

FREE

When status is success, result is VideoAnalysis (default) or a workflow format (recipe, creator-brief, workout-plan, etc.) when format was set on submit.

RESPONSE.JSONjson
{
  "platform": "tiktok",
  "url": "https://www.tiktok.com/@janecreates/video/7123456789",
  "title": "How I grew my account in 30 days",
  "author": "Jane Creator",
  "authorHandle": "@janecreates",
  "transcription": "Here's how I grew my account in 30 days...",
  "language": "en",
  "onScreenText": "Day 1 · Follow for more",
  "visualDescription": "Creator speaking to camera in a home office, cuts to screen recording of analytics dashboard.",
  "scenes": [
    {
      "timestamp": "0:00",
      "description": "Talking head, direct to camera"
    },
    {
      "timestamp": "0:08",
      "description": "Screen recording of analytics dashboard"
    }
  ],
  "processedAt": "2026-05-20T12:34:56Z"
}

// §05 · AGENTIC

Built for
autonomous agents.

No accounts. No API keys. No subscription. Agents authenticate by paying. Pick any of 24 workflow formats for typed outputs — every shape has a published JSON Schema. Cached repeat lookups return instantly.