A proven taxonomy, in the pipeline you already run.
YAHO judges every scene against the Pyler taxonomy, proven on over a billion videos. Send video the way you already store it, get JSON back, and use it across your systems.
Verdicts from a taxonomy proven on over a billion videos.
The Pyler taxonomy is our own set of categories, used to judge over a billion videos and backed by a dataset labeled to the GARM brand safety framework. YAHO applies it to every scene, in content and ad creatives alike.
- 1B+videos judged with the Pyler taxonomy
- GARMlabeled dataset built to the brand safety framework
Pyler taxonomy
Our own categories, used to judge over a billion YouTube videos, backed by a dataset labeled to the GARM brand safety framework.
- Controversy
- Hate
- Politics
- Sexual content
- Violence
- + by country
taxonomy: pyler proven on: 1B+ YouTube videos (video-level) labels: GARM brand safety framework applied: every scene, in videos and ad creatives
Policy controls
Start from the Pyler taxonomy and add your own categories and lists.
- Custom categories
- Allow and block lists
- Lists apply in order: global, local, then organization
policy: nova-tv-v3 categories: pyler default custom: "Disaster & accident" block list: 1,204 channels allow list: 312 titles
Verdicts with reasons
Every scene gets one of three outcomes, with a category and a reason. Use them to apply content policies or ad rules.
- Allow · block · review
- A category and a reason on every verdict
- Timecoded, so you act on the exact scene
{
"scene": 4,
"start": "01:05", "end": "01:30",
"verdict": {
"decision": "review",
"category": "Disaster & accident (custom)",
"reason": "Vessel at risk of capsizing"
}
}Every verdict sits on scene data you can use too.
Every video is split into scenes, and each scene is described in a structure that people, ad systems and AI agents can read. Use it for search, catalogs and recommendations too.
Scene Detection
Splits every video into scenes and finds what is in each one.
- Scene boundaries with timecodes
- Objects and people in each scene
- On-screen text and captions
- Speech and sound events
SC 01 00:00–00:22 harbor · lighthouse SC 02 00:22–00:41 open water · wind SC 03 00:41–01:05 storm clouds · waves SC 04 01:05–01:30 sailboat · breaking waves audio: wind · rain text: none
Scene Analysis
Describes each scene and the whole video, mapped to IAB Content Taxonomy 3.0. Ad creatives are analyzed scene by scene too.
- Whole-video summary and mood
- A description for every scene
- IAB Content Taxonomy 3.0 per scene
- Ad creatives scene by scene, for matching and blocking
- Metadata for catalogs, recommendations and highlights
video: "A sailing documentary that moves
from a calm harbor into a storm."
mood: calm → tense → dramatic
scene 4: "A small vessel tilts hard in
breaking waves."
iab: Travel › SailingScene Search
Describe the scene you need in everyday language and search the library.
- Search the whole library
- Timecoded scenes, ranked
- Across every analyzed video
query: "boat in rough sea" 1 open_sea.mp4 01:05–01:30 0.93 2 open_sea.mp4 00:41–01:05 0.88 3 harbor_doc.mp4 12:10–12:34 0.81
Plug it into the pipeline you already run.
Send video the way you already store it and get one standard JSON back. The Console and the API share one quota.
- IN
Send what you have
- File upload · URL
- MRSS feed
- Live stream
- Ad creatives
- OUT
Get one standard output
- Scenes with timestamps
- Description · IAB 3.0
- Pyler category · verdict · reason
JSON per scene · VOD within 30 minutes
- TO
Receive it where you work
- REST APIInto your systems
- Web ConsoleReview results, manage policy
Three calls, from video to verdicts
POST /v1/videos
# 1. Send a video URL (202 Accepted — indexing runs asynchronously)
curl -X POST https://video.pyler.ai/api/v1/videos \
-H "Authorization: Bearer $YAHO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"url": "https://cdn.example.com/open_sea.mp4", "title": "Open Sea"}'
{ "video_id": "7c1f4a2e-…", "job_id": "9b0d6e51-…", "created": true }Free trial: up to 120 minutes per video, 5 files per upload, 30 ad creatives.
One output, many places to use it.
The same scene data and verdicts feed your ad server, SSAI, CMS and review tools. Your systems decide what to do with them.
- Content video
- Ad creatives
VIA: File · URL · MRSS · Live stream
Understanding layer
- Scene detection
- Scene analysis
- Scene search
- Metadata
Judgment layer
- Pyler taxonomy
- Policy
- Verdict
Your policy: custom categories + allow / block lists
- Ad server · SSP
- SSAI · playout
- CMS · catalog · search
- Review tools
DELIVERED THROUGH: Web Console · REST API