VIDTREO Intelligence

VIDTREO Intelligence turns recordings into structured data you can act on. You describe once what matters — in a lens — and every video or audio recorded in your environments is analyzed automatically: a written summary, the data points you asked for, a score against your rubric, and the speakers, all tied to the exact moments of the recording that support them.

Results reach your systems through the intelligence.lens_analysis.completed webhook, together with the userMetadata you attached when recording, so they flow into your ATS, CRM, LMS, or helpdesk without a single API call.

Recording completes (video or audio)
    → Each lens active in the environment queues an analysis
    → Transcript and speakers (+ visual review when the lens checks the screen)
    → The lens is applied: summary, data points, scores — every claim cites its evidence
    → VIDTREO validates each citation and computes the scores
    → intelligence.lens_analysis.completed webhook fires with the full result

Key Concepts

What You Get

For each recording and each lens, the result contains:

• Summary — an overall summary plus the sections you defined, as labeled points
• Data points — typed values extracted from the recording (findings), each with a status: supported, insufficient evidence, or contradictory
• Scorecard — a score per rubric criterion with written feedback, and a total computed by VIDTREO
• Speakers — who speaks, their role, and their name when it can be established
• Chapters — chronological navigation points with timestamps
• Quality checks — counts of citations, invalid citations, unsupported items, and abstentions

The full structure is documented in the Result reference.

Enabling Intelligence

Intelligence is available to every VIDTREO account. Turning it on for an environment takes four steps in the dashboard:

From then on, every recording that completes in the environment is analyzed by each active lens with auto-analyze on, and each completed analysis is delivered to your endpoint.

Integration at a Glance

For developers, a complete integration has two touch points: attach your identifiers when recording, and handle the webhook.

1. Attach your own identifiers to each recording with userMetadata:

<VidtreoRecorder
  apiKey="vt_live_xxxxxxxxxxxxxxxxxxxx"
  userMetadata={{ candidateId: 'cand_8231', jobId: 'job_114' }}
/>

2. Receive the analysis — your identifiers come back in the same payload as the result:

app.post('/webhooks/vidtreo', express.raw({ type: 'application/json' }), async (req, res) => {
  // Verify X-Webhook-Signature over the raw body first (see Results & Webhook)
  const event = JSON.parse(req.body.toString('utf8'))
  if (event.event !== 'intelligence.lens_analysis.completed') return res.status(200).end()

  const { run, result } = event.data
  const recording = event.data.video ?? event.data.audio
  await updateCandidate(recording.userMetadata?.candidateId, {
    summary: run.headline,
    score: result.scorecard?.total ?? null,
    findings: result.findings,
  })
  res.status(200).end()
})

The Results & Webhook page has the full payload, the TypeScript types, and production-ready handlers in Node.js and Python.

Analyzing a Recording Manually

Use manual analysis to test a lens before turning on auto-analyze, to analyze recordings made before the lens was active, or to apply a lens that is not active in the environment.

You can also start from the lens itself with Analyze a recording on the lens page.

Automatic analyses always use Spanish and English, with no question and no participants.

Rules worth knowing:

• The recording must have finished processing: The video is still processing; analyze it once it is ready.
• Requesting the same lens version again while an analysis of that recording is still in progress returns the in-progress analysis instead of starting — and charging — a second one (unless you provide a question or participants).
• Analyzing a recording again with the same lens creates a new analysis with its own result, charge, and webhook.
• Transcripts are reused: when a recording is analyzed by several lenses, it is transcribed once.

How an Analysis Runs

Each analysis moves through these steps, visible live in the dashboard:

Statuses

Common failure reasons:

Restart a failed analysis from the recording page with Try again.

Pricing

Every analysis is charged on its own from your prepaid credits, the same way recordings are.

• Minutes are rounded up, with a one-minute minimum, based on the recording's duration.
• Each lens is a separate analysis: a recording analyzed by three lenses is charged three times.
• Re-analyses are charged in full. Failed analyses are free.
• Current prices are always shown in Billing → Pricing, under Vidtreo Intelligence.

How charging works:

1. Hold — when an analysis is requested, its price is reserved from your balance. If the balance cannot cover it (after auto-recharge, if enabled), a manual analysis is refused with Not enough credits, and an automatic one is recorded as failed with the reason.
2. Charge or release — the hold is charged when the analysis completes and released if it fails. Holds still open after 24 hours are released automatically.

Each analysis shows its charge in the dashboard ($X on hold, $X, or Not charged), and charged analyses appear in Billing → Billing history as Intelligence analysis.

Permissions

Users only see lenses, analyses, and results for the environments they have access to.

In the Dashboard

Beyond the webhook, every analysis can be reviewed in the dashboard:

• On the recording page — the analysis bar under the player lists each analysis with its lens, score, status, and charge. The report shows the summary, data points, scorecard, and coverage; the side panel shows Chapters, Transcript, Evidence (play every cited moment as a reel), and Ask to question the recording in natural language, with answers that cite timestamps.
• On the lens page — Overview, History (compare versions), Where it runs (activations), and Analyzed recordings.
• Ask your recordings — from Intelligence, ask questions across every analyzed video and audio; answers cite the exact moments and list the recordings consulted.

Next Steps