50,000 hours of video, searchable in milliseconds.
Meshora watches your archive — indexing every spoken word, on-screen text, visual scene, and emotion — so your team finds any 14-second clip without touching a spreadsheet.
Petabytes of video nobody can actually search.
Your archive staff spends 4 hours finding a 40-second clip. Title tags and manual metadata cover maybe 3% of what's in your footage. The rest is locked inside the frames, unindexed, unfindable.
Three steps from upload to search.
Connect your storage
Point Meshora at your S3 bucket, GCS, Azure Blob, or upload directly. We handle the rest.
Multimodal analysis
Every frame analyzed. Speech transcribed. Text read. Scenes classified. Entities linked. Built into a searchable knowledge graph.
Find any moment
Natural language queries return timestamped results with transcript excerpts and confidence scores. API or dashboard.
Four indexing layers. One search interface.
Speech Index
94% transcript accuracy across English and 12 languages. Every word spoken in your archive becomes a searchable token with precise timecode.
Visual Scene Detection
14 scene-level signals tracked per frame: visual complexity, motion, speaker presence, emotional tone, and more. Find the moment by what it looks like.
OCR Text Extraction
Ticker text, chyrons, overlaid captions, slide content, signage — all indexed. The most ignored signal in video search, now fully captured.
Knowledge Graph
Entities linked across clips: the same expert, the same product name, the same location — surfaced as connections across your entire archive.
Different archive. Same broken search problem.
Search 40 years of broadcast footage in seconds.
Regional broadcasters and documentary makers hold decades of footage. Most of it sits unreachable because the metadata was typed by hand and covers about 3% of what's there. Meshora indexes the other 97%.
- Search spoken words across archived news broadcasts
- Find footage by on-screen text — ticker text, chyrons, graphics
- Locate archival clips 3× faster for documentary re-use
Make your training library actually usable.
Corporate training videos collect dust because employees can't find what they need inside them. Meshora turns your onboarding and procedure library into a searchable database — find the exact procedure in 90 seconds, not 8 hours.
- Search across all onboarding and compliance videos
- Link related procedures across departments via knowledge graph
- Cut new-hire onboarding from 8 hours to 90 minutes
Find the statement without watching the deposition.
Discovery reviewers watch entire depositions looking for one sentence. Meshora indexes every spoken word in your deposition recordings — search for the statement, get the timestamp, skip the linear review.
- Full transcript search across deposition recordings
- Visual evidence log: on-screen documents and exhibits indexed
- Reduce discovery review hours by up to 70%
Real results from real archives.
"We uploaded four years of regional news footage. Meshora had it searchable before lunch. Our researchers were finding clips they'd given up looking for."
"Our compliance team reviews deposition recordings for specific statements. What used to take a paralegal three days now takes twenty minutes with a search query."
Built for developers who need to ship.
REST API with Python and Node.js SDKs. Webhook callbacks when indexing completes. JSON responses with segment-level results, confidence scores, and transcript excerpts. Start querying in under 30 minutes.
- POST /v1/index — submit video for indexing
- POST /v1/search — natural language queries
- GET /v1/export/:id — clips, transcripts, metadata
import meshora
client = meshora.Client("sk_live_...")
# Search 50,000 hours in milliseconds
results = client.search(
query="product recall announcement March 2024",
archive_id="arch_q1_broadcast"
)
for clip in results.clips:
print(f"{clip.timecode} — {clip.confidence:.0%}")
# 01:24:07 — 96%
# 00:04:33 — 89%
Index your first archive free.
Start with 10 free hours. No credit card. See what's been hiding in your video archive since you uploaded it.