Introducing VIDTREO Intelligence: answers from every recording, with the receipts

TL;DR

• VIDTREO Intelligence is live. Describe what you need to know from your recordings. Intelligence returns structured data, a summary by section and a scorecard for every video and audio.
• Every answer shows its proof. Points, data points and scores cite a moment of the recording, and Intelligence checks each citation against it. When the recording doesn’t support an answer, it abstains instead of guessing.
• It runs on its own. Turn a lens on for an environment and every new recording is analyzed. Results reach your system as JSON through the intelligence.lens_analysis.completed webhook.
• $0.035/min for video, $0.025/min for audio. Paid from the same prepaid credits as your recordings. Failed analyses are never charged.

A transcript is not an answer

In August, VIDTREO AI started transcribing every recording automatically. Video became text. Text is not a decision.

Someone still has to read the transcript, decide what it means and copy the result into the CRM, the ATS or the QA sheet. A summary shortens the reading. It doesn’t make the call, and it doesn’t tell you where its conclusions came from.

An AI score is worse. A “7/10” with no criteria and no evidence is a number nobody can defend in front of a candidate, a customer or an auditor.

We built Intelligence for decisions people have to defend. A week ago we announced it was in its final testing stages. Today you can use it.

A lens is what you want to know, written once

Intelligence works with lenses. A lens holds the objective, the questions and the rules of an analysis. Run it on one recording or on every new one, and each result comes back in the same shape.

A lens has six parts:

• Objective: what the analysis is for, in one sentence.
• Summary sections: the headings every summary must cover.
• Data points: text, numbers, yes/no, choices and lists your system can read.
• Scorecard: criteria with a maximum score, for rubrics like interviews or QA.
• Trust rules: what the analysis must never assume or invent.
• On-screen checks: yes/no questions about what appears in the video.

Here is a lens for sales discovery calls:

Lens: Sales discovery call

Objective         Find out if the prospect is ready to buy and what is blocking them.
Summary sections  Pain points · Next steps
Data points       Budget confirmed (yes/no) · Decision maker (text) · Objections (list)
Scorecard         Discovery questions · /5
Trust rule        Never assume a budget that wasn't said out loud.
On-screen check   Is a pricing slide on screen?

The same structure works for an interview answer, a support call, a class or a compliance review. Only the content of the lens changes.

You don’t write prompts. The Lens Designer does

Building a lens doesn’t require prompt engineering. Drop in a recording or describe what you’re looking for, and the Lens Designer drafts it with you:

1. Start from a recording or a sentence. Upload a video or audio, pick one you already have or describe what to detect.
2. It reads before it suggests. The Designer writes down what was said and who said it, then proposes two or three lenses based on what is actually in the recording.
3. Answer with a tap. Short rounds of questions, each with a recommended option. Never a blank form.
4. Publish and analyze. You see the cost before you run it. Every later change is saved as a new draft with a note.

You can still edit every part of the lens by hand. Try the whole flow on a sample recording:

Sample

Vidtreo Intelligence

What should we detect in your recordings?

Drop a video or audio recording. The Designer reads it and asks you a few quick questions — just pick the answers.

Drop a video or audio file here, or browse

In this demo, use the sample recording below.

Discovery call · AcmeVideo · 33:10Use this recording

Once the lens exists, you keep shaping it by talking to the Designer. Accept one of its suggestions or ask for a change, and watch the draft update. Everything the Designer adds is marked, and every change is saved as a new draft:

Sample

Sales discovery call · from v1

Back to lens

01

Goal

What every analysis tries to answer.

Find out if the prospect is ready to buy and what is blocking them.

02

What each analysis includes

The parts of the report you get for every recording.

Summary

Sectioned summary of the analyzed recording.

Summary

Data points

Structured field values extracted from the recording.

Findings

Discovery questions

Rubric-based score across weighted criteria.

Scorecard

Chapters

Timestamped chapter breakdown of the recording.

Chapters

Evidence

Timestamped quotes backing every claim.

Evidence

Coverage

Support, abstention, and contradiction report.

Coverage

03

Summary sections

The headings of the written summary.

Pain pointsNext steps

Add

04

Data points

Specific answers pulled from every recording, ready to filter and export.

Key	Type	Description	Remove
Budget confirmed	boolean	Did the prospect confirm a budget out loud?
Decision maker	text	Who makes the final decision, with their role.
Objections	list of text	Every objection the prospect raised.
Competitors evaluated	list of text	Competitors the prospect names.

05

Scoring (out of 5)

How the recording is scored, criterion by criterion.

• Current process2 pts
• Budget and timing2 pts
• Decision process1 pt

06

Rules it follows

What the analysis must never assume or invent.

• Never assume a budget that wasn't said out loud.

Add

What comes back

Each analysis returns an Intelligence Package: a headline, a summary by section, chapters, data points, a scorecard and the evidence behind each of them.

Every recording goes through five steps:

1. Transcript and speakers: what was said, split by speaker.
2. Visual review: frames checked against the lens’s on-screen questions.
3. Who is speaking: names matched from introductions and video-call labels.
4. Lens analysis: your objective, data points and scorecard, applied.
5. Evidence checks: every citation validated against the recording.

This is the result for the discovery call above, as it looks in the VIDTREO Dashboard. Click any timestamp and the recording jumps to the moment that proves it. Open the Evidence tab and press Play as a reel to watch every cited moment back to back.

Sample

DWDana Whitfield

ARAlex Rivera

Thanks for making time, Dana. How does your team review calls today?

0:00 / 33:10Current process

Dana Whitfield47%

Alex Rivera53%

ChaptersTranscriptEvidenceAsk

6 cited moments

Play as a reel

1. 0:45Pain pointsDana Whitfield“Today three people watch every onboarding call and fill a spreadsheet by hand.”
2. 1:47Decision makerDana Whitfield“I run operations, so the final call on tools like this is mine.”
3. 4:12Budget confirmedDana Whitfield“We set aside budget for this in Q3.”
4. 9:30ObjectionsDana Whitfield“Our IT team will need a security review, and onboarding can't take more than two weeks.”
5. 27:00Next stepsAlex Rivera“We can start with one team and measure review time for a month.”
6. 31:20Next stepsDana Whitfield“Let's book the security review with our IT lead next Tuesday.”

SummaryData pointsDiscovery questionsCoverage

Ready to buy in Q3 if the security review passes.

Dana confirmed the budget and owns the decision. A security review and a two-week onboarding limit are the only blockers.

Pain points1

• Three people review every onboarding call by hand and copy the results into a spreadsheet.
  0:45“Today three people watch every onboarding call and fill a spreadsheet by hand.”

Next steps2

• Security review call with Acme's IT lead next Tuesday.
  31:20“Let's book the security review with our IT lead next Tuesday.”
• Start with one team and measure review time for a month.
  27:00“We can start with one team and measure review time for a month.”

Built to be checked, not just believed

Summarizing a video is the easy part. What matters is what happens when the model is wrong or the recording doesn’t have the answer.

• Every claim is cited. Points, data points and scores must cite a moment of the transcript or a frame. Intelligence validates each citation against the recording and discards the ones that don’t match.
• No evidence, no score. A criterion whose evidence doesn’t hold up loses its score and is marked as insufficient evidence.
• It abstains instead of guessing. When the recording doesn’t support an answer, the result says so. Nobody mentioned competitors on the call above, so that data point stays empty.
• Contradictions are flagged. If the recording says two different things, the data point is marked as contradictory.
• The total is the sum of its criteria. If any criterion lacks evidence, the scorecard is marked incomplete and there is no total. You never get a subtotal dressed up as a final score.
• No emotion reading, no face recognition. The visual review never infers emotions or personality and never identifies people by their face.
• Your reviewers make it better. Mark any claim right or wrong. Every review joins the lens’s gold set, and accuracy is tracked for each version.

Ask your recordings anything

Sometimes you don’t need a lens, you need an answer. Ask lets you chat with one recording from its player, or across every video and audio you’ve analyzed.

Try it on the same sample call. Pick a suggestion or type your own question: every citation jumps to its moment in the recording.

Sample

DWDana Whitfield

ARAlex Rivera

Thanks for making time, Dana. How does your team review calls today?

0:00 / 33:10Current process

Dana Whitfield47%

Alex Rivera53%

ChaptersTranscriptEvidenceAsk

Ask anything about this video

Answers come from what was said and shown, with the exact moments.

What are the key takeaways?When does the most important moment happen?What was left unresolved?

Answers come only from what was said and shown, with the exact moments. When the answer isn’t in your recordings, Ask tells you. It replies in the language you ask in.

Runs on every new recording

Publish a lens, turn it on for an environment, and every video or audio that arrives there is analyzed on its own. You choose where each lens runs and whether automatic analysis is on or off.

Lenses are versioned: drafts, published versions, change notes and a comparison between any two versions.

When an analysis finishes, your endpoint receives the full result. Trimmed, it looks like this:

{
  "event": "intelligence.lens_analysis.completed",
  "data": {
    "run": {
      "lensName": "Sales discovery call",
      "mediaKind": "video",
      "headline": "Ready to buy in Q3 if the security review passes.",
      "score": "4/5"
    },
    "result": {
      "findings": [
        {
          "field": "Decision maker",
          "value": "Dana, VP of Operations",
          "status": "supported",
          "evidence": [{ "kind": "segment", "start": 107.2, "end": 112.8 }]
        },
        {
          "field": "Competitors evaluated",
          "value": null,
          "status": "insufficient_evidence",
          "evidence": []
        }
      ]
    }
  }
}

Every finding carries its status and the exact seconds that support it. score is null when the scorecard is incomplete, so your code never mistakes a subtotal for a final score.

Intelligence works with any video or audio in your VIDTREO library, whether it was recorded with VIDTREO Recorder or uploaded.

Pay per analyzed minute

Recording	Price per minute, per analysis
Video	$0.035
Audio	$0.025

• A 10-minute video costs $0.35 per lens. Two lenses on the same video cost $0.70.
• Each analysis is rounded up to the next minute.
• Failed analyses are never charged.
• No seats, no monthly fee: analyses use the same prepaid credits as your recordings.

Recording with VIDTREO Recorder too? HD video costs $0.010/min with storage and transcription included, so recording and analyzing a minute of video costs $0.045. Your $1 free credit covers about 28 minutes of analyzed video. Estimate your own volume with the Intelligence calculator.

Languages and media

• Intelligence understands Spanish and English, including recordings that mix both. Results can be written in Spanish, in English or in the main language of the recording.
• The visual review only runs on video, when the lens has on-screen checks. Audio gets the transcript, the speakers and the full lens analysis.

Get started

1. Create your account and get $1 in free credit.
2. Open the Lens Designer in the VIDTREO Dashboard and describe what you need to know, or drop in a recording.
3. Publish the lens, turn it on for an environment and subscribe to intelligence.lens_analysis.completed.

Your recordings already hold the answers. Now every answer comes with its proof.

Start free — $1 credit → See how Intelligence works → See full pricing →