In beta

    Ask your focus groups anything.
    Every answer cites the exact minute.

    Bring five hours of sessions or forty. Finding the moment and quoting it is mechanical work; what it means for the brand is still yours to decide.

    The problem

    Forty hours of video, and the quote you need is inside one of them.

    A qualitative project ends with five to forty hours of focus groups, triads and interviews. Three ways out exist today, and every one of them costs you something.

    Watch it all again

    The only way to be certain, and the reason a project takes a week longer than it should.

    Pay for a transcript

    You get the words. You lose the voice, the face, and whatever was on the screen.

    Reach for a product research tool

    It does not know that a session has a moderator, that a commercial was played, or that the answer was printed on a pack.

    How you use it

    From a folder of recordings to an answer with its citation.

    Six steps, and you see what each one costs before it runs.

    1. 1

      Drag your files in

      The upload resumes if your connection drops, and you watch the speed and the time left while it runs.

    2. 2

      See the cost before you spend it

      You see the length of every file and the total before anything is processed. If your hours do not cover it, nothing starts.

    3. 3

      Transcription with the voices kept apart

      Every word carries its own timestamp, and each speaker is separated from the others.

    4. 4

      Roles, not just speakers

      The moderator is told apart from the participants. A commercial played inside the session is marked as a played stimulus instead of being credited to a person.

    5. 5

      What was on the screen

      Frames are read every fifteen seconds: the scene, the people, the brands, and the on-screen text word for word.

    6. 6

      Ask

      The answer arrives with numbered citations. Click one and the video jumps to that second.

    What was measured

    99.3% of the citations are verbatim.

    On 25 September 2026 the system answered 37 questions across 11 videos. All 275 citations it produced were checked against a second, independent transcription made by a different engine. 273 matched word for word.

    The automatic comparison cleared 94.9%. The 14 cases it flagged were read by hand, and 12 of them turned out to be differences between the two transcriptions rather than invented quotes. The other 2 are the 0.7%.

    Coverage is a different question from accuracy. On the twelve questions put to a four-interview collection, the system drew on all four relevant sessions every time.

    Where it does worse

    The same test found three things it does not do well.

    In a video that mixed Spanish and Portuguese, the transcription turned a Portuguese passage into Spanish. A citation taken from there is faithful in meaning and not literal in wording.

    On heavily edited video, and on questions whose answer exists only in the image, it performs worse than on a plainly recorded session. The visual collection was the weakest part of the test.

    The image analysis can fail on some stretches. The dialogue stays available, and half the time for those stretches is returned to you.

    How it is sold

    Hours, not a subscription.

    You buy hours and spend them when a project needs them. They do not expire, the same as the rest of the suite, and there is no seat licence, no sales call to sit through, and no renewal to remember to cancel.

    A new account gets one free hour, once.

    Buying is not open yet. During the beta, hours are granted by invitation.

    Billing that gives the time back

    • Hours are reserved when you confirm the batch, not when you upload it.
    • If a video fails, its hours come back to you.
    • If the real length differs from the estimate by more than a minute, the charge is adjusted.
    • If only the image analysis fails, half the time for those stretches is returned and the dialogue stays available.
    The limits

    What fits, and how much you can ask.

    Up to 5 GB per file and 8 hours per video.

    Up to 20 files in a single batch.

    mp4, mov, mkv, webm and m4v.

    Fifty questions per processed hour, per project.

    Not built yet

    What comes next, and does not exist today.

    None of these three works today. They are here so you know where the product is heading, not so you plan a project around them.

    Import from Google Drive

    Bring recordings straight from a Drive folder instead of downloading them and uploading them again.

    Share an answer with clips

    Send a client the answer together with the video fragments that support it, on a page of its own.

    A report built from your objectives

    Upload the objectives of the study and receive a first draft of the report, organised by those objectives.

    How it works

    From video to a citation you can check yourself.

    Retrieval over video, built for market research rather than for meetings. This page is the pipeline stage by stage, with what it gets right and where it fails.

    It is in beta, with invited researchers.

    The product works and it is not open to the public yet. Tell us about your project and we will write to you when a place opens.

    Ask about the beta