From raw footage to searchable archive in five steps.

No manual tagging. No spreadsheets. Connect your storage, and Meshora indexes every spoken word, on-screen text, visual scene, and entity connection automatically.

Start in 5 minutes
01 Ingest
02 Analyze
03 Graph
04 Query
05 Export

Connect your storage. Point and index.

Meshora connects to your existing file store in minutes. There's no proprietary format, no re-encoding step, and no file size limit. We handle the queue management and retry logic for large archives.

Amazon S3 via IAM role
Google Cloud Storage service account
Azure Blob Storage SAS token
Direct upload via dashboard or API

Supported formats: MP4, MOV, MKV, AVI, MTS, MPEG-2, ProRes. Maximum file size: 50GB per file.

Four signal streams running in parallel.

Meshora processes every file through four analysis pipelines simultaneously. By the time one stage completes, the next is already warm. A one-hour video takes roughly 18 minutes end-to-end.

Speech transcription

94% word accuracy across 13 languages, with speaker diarization. Every word timestamped to the nearest 100ms.

Scene detection

14 visual signals per key frame: motion level, speaker presence, emotional tone, object classes, scene type.

OCR extraction

All text that appears on screen — chyrons, ticker, slides, signage — extracted and timestamped for search.

Sentiment + emotion

Tone classification per speech segment: neutral, positive, tense, formal. Useful for compliance and editorial review. (Pro and Archive tiers.)

Entities linked across your entire archive.

After analysis, Meshora runs entity extraction and cross-clip linking. Named people, places, organizations, and products become nodes in a graph. The same expert who appears in 400 training clips gets recognized as the same entity — not 400 separate mentions.

  • Person recognition across clips — no account required, no face matching stored
  • Organization + product name entity linking from speech and OCR combined
  • Query entity connections — "show all clips where this product name appears"
  • Graph exported via API — use it in your own applications

Natural language. Sub-50ms results.

Once indexed, your archive is queryable via the Meshora dashboard or REST API. Type a question in plain English. The system returns timestamped clips ranked by relevance, with a short transcript excerpt and the specific signals that matched.

spokesperson mentions product recall March 2024 47ms
01:24:07 96% broadcast_archive_Q1_2024.mp4 — "...we are initiating a recall of all units manufactured before March 2024..."
00:04:33 89% press_conf_2024-03-14.mp4 — "The product recall announcement covers three product lines..."

Take the data where you need it.

Meshora is not a walled garden. Every index, transcript, and result is available via API. Export full transcripts as JSON or SRT. Download clip segments as MP4. Pull the knowledge graph as a node-edge manifest.

Full transcript as JSON with per-word timestamps
SRT subtitle file export from transcripts
Clip segments trimmed to query match windows
Knowledge graph as JSON edge list

Start indexing in under 5 minutes.

Connect your S3, GCS, or Azure Blob — or upload directly. Drop in your first batch. A one-hour video is searchable in about 18 minutes. 10 free hours, no credit card.