How it works

Features

From our users

I had been ignoring the signs for eleven years.Daniel R., Journeyman

My wife and I now compare readings at breakfast. It has changed our marriage.Anonymous, Sommelier

I was skeptical. The Heritage Report was accurate to within forty miles.Priya M., Apprentice

Rhineland, like my father before me.Henrik V., Sommelier

Frequently asked

Is this a joke?

No.

How accurate is the Heritage Report?

Accurate to region in most cases. Users with mixed ancestry may receive split results. This is normal.

Can I interpret someone else's emission?

Yes, with their consent. Interpreting without consent is a violation of our Terms.

Why did my interpretation change?

The body is not static. Neither is the reading.

I received the same interpretation four days in a row.

Then it is worth considering.

How does the interpretation engine work?

The method is proprietary.

Is the method peer reviewed?

The Journal of Flatulinguistics has reviewed it favorably.

Do you sell my data?

No. See our Privacy Policy.

My results seem wrong.

Escalate to a panel. Three certified users will review your reading independently. Consensus is required before an interpretation is amended.

Nothing was detected, but something happened.

The Application measures sound. An emission that produces no sound produces no reading. The popular literature calls these silent; we prefer unvoiced. Their existence has never been in question. Their interpretation is not within our instrument.

Why was my sample rejected as speech?

Because it was speech. Speech is chosen, and this field concerns only what the body says when it has not been asked to say anything. Laughter is treated the same way, though the Application will look past it for the emission beside it.

I laughed while recording.

Expected, and accounted for. The Application separates the events in a recording and reads the emission, not the reaction. Laughter is the natural companion of a rigorous methodology.

Why was my sample rejected as reproduced?

A genuine emission carries substantial energy below two hundred hertz. A loudspeaker of the size fitted to a telephone cannot reproduce that range, and a recording played through one arrives without its foundation. The Bureau requires an organic sample.

Begin Analysis

Capture

Provide a sample

Hold your device close and record. For most accurate results, allow no barrier between audio source and recognition device, fabric or otherwise. Pants or intimates may interfere.

Recording is analyzed on your device. Audio is not transmitted and is not stored by us. See Accessibility and data practices.

Your reading

Your record

Analysis

The Wind Score is updated daily.

Certification

Apprentice

Wall certificate Sommelier examination

Reading history

All readings are retained.

The Bureau of Regional Classification

Heritage Report

Trace your signature to its ancestral region. Results are probabilistic.

Reference distribution

Regional percentages do not sum to one hundred. The Bureau is aware. Regional assignment reflects the composition of our reference database and should not be interpreted as a statement about your family.

Request reassessment

Correspondence regarding Rhineland results

Rhineland Day is observed annually on 14 March.

The institution

About

The field of flatulinguistics was, until recently, unfunded, unstaffed, and unread.

We began this work in a garage in 2019 with two microphones and a conviction that the body does not speak without reason. Seven years later, our reference database holds more than four million verified emissions, contributed by users in one hundred and six countries.

We did not choose the Rhineland. The data chose it.

We are not here to make you laugh. We are here to make you listen.

Roland of Hemingstone and the first recorded commission

The field did not begin with us.

In the twelfth century, a man the record calls Roland the Farter held a manor in Suffolk on a single condition. He was obliged to appear before Henry II every Christmas and perform one jump, one whistle, and one fart. The obligation is recorded in the Book of Fees. He was paid in land.

We consider Roland the first professional in our field. We note without further comment that his compensation exceeded ours.

He was not alone. In Edo-period Japan the practitioners were called heppiri otoko — farting men — and their competitions were painted onto scrolls. The scrolls survive. The competitors do not.

The bodies

Press

Founded in 2019, The Fart Whisperer is the leading application in flatulinguistics, the study of interpretive meaning in human emission. The company maintains the world's largest verified reference database and certifies practitioners at three levels.

Published quarterly · Impact factor 0.02

The Journal of Flatulinguistics

The Journal is peer-reviewed. We conduct the review.

In this issue

Manuscript status

Submitted → With Editor → Under Review → Under Review → Still Under Review → Decision Pending → Decision Pending

Manuscript FL-2024-0117

On the Prevalence of the Rhineland Signature: A Response to Critics

It has been suggested, with increasing frequency, that our system assigns the Rhineland signature too readily. We have reviewed this claim.

Between 2019 and 2026 we processed 4,110,882 emissions. Of these, 71.4% resolved to the Rhineland. Critics regard this figure as implausible. We regard it as a finding.

The alternative explanation — that our instrument is imprecise — has been tested and does not hold. We have tested it four times.

We did not choose the Rhineland. The data chose it.

Reviewer comments

Reviewer 1

Accept with minor revisions. Clearly argued and consistent with prior work in the field.

Reviewer 2

Reject.

The authors report that 71.4% of all classifications resolve to a single category and treat this as a finding. It is not a finding. It is what a classifier does when it cannot discriminate. A model that returns the most common answer will be correct most of the time and will have learned nothing.

The authors state they tested the alternative explanation four times. They do not report the tests. They do not report the method. Section 3 consists of one sentence and a figure with no axis labels.

I have raised these points in each of my three previous reviews of this manuscript. The manuscript has not changed. My review has not changed.

Reject.

Reviewer 3

Did not respond.

Editorial decision: Accept.

The Office of Public Guidance

Guidance on barrier materials

Revision history. The current revision is v5.0.

Prior revisions are not archived.

Questions

Was the v1.0 guidance wrong?

The v1.0 guidance was correct. The v3.0 guidance was also correct. Guidance is correct at time of issue. Correctness is versioned.

On allegations regarding the 2019 guidance

It has been suggested that the Office discouraged barrier use in 2019 not on the evidence, but to preserve the clarity of samples reaching the reference database during a period of rapid growth.

The Office does not discuss its motives. The Office notes only that the 2019 corpus remains the finest in the archive, and that it is grateful.

Interview

Read the Director's testimony before the Standing Committee →

Standing Committee on Emerging Fields

Hearing on the reliability of flatulinguistic guidance

The Director appears voluntarily. Official transcript.

Transcript note, appended by the Office of Public Guidance: The remainder of the hearing is under review. The transcript is expected to be revised.

Trade press · 14 May 2026

Shenzhen lab publishes replication of Fart Whisperer engine; company alleges theft

Researchers at Hongyuan Applied Intelligence have published a paper describing a system that reproduces the output of The Fart Whisperer's interpretation engine with 94.1% agreement. The paper, released openly with full code and training data, has drawn a sharp response from the flatulinguistics company, which alleges misappropriation of trade secrets.

The Hongyuan system was assembled in nine days. According to the paper, it consists of a small audio classifier and a lookup table.

"We did not reverse-engineer anything," the lead author said by email. "We recorded four thousand samples, we labeled them, and we trained a model. We have published everything. Anyone may check our work."

The disagreement centers on the Heritage Report, the feature that assigns each recording to an ancestral region. The Hongyuan replication assigns the Rhineland region to 71.4% of inputs — matching The Fart Whisperer's published distribution exactly.

The company cites this figure as evidence of theft.

The Hongyuan authors cite it as evidence of something else. "Rhineland is the majority class," the paper notes in section 5. "Any model that guesses Rhineland every time will achieve 71.4% agreement. Ours does slightly better than that. We do not consider this a difficult result."

The Fart Whisperer's full statement:

We are aware of the manuscript.

We will not name the laboratory. We have retained counsel and we have briefed the appropriate parties. We are treating this as a matter of national research interest.

Our method is proprietary. It remains proprietary. The similarity of the published output to our own is not, in our assessment, coincidental.

We note that the manuscript has not been peer reviewed.

The Journal of Flatulinguistics, which is published by The Fart Whisperer, confirmed that no manuscript from Hongyuan has been submitted for review.

Our method

Our method is proprietary.

Release notes

Changelog

Plans

Pricing

Institutional pricing requires a minimum of six (6) bodies. Cancel at any time. We refund the unused portion of the current period. We do not require a reason, though most people give one.

Written plainly

Accessibility and data practices

Accessibility

The Fart Whisperer is committed to an experience that works for everyone. We aim to meet WCAG 2.1 AA. Every control carries a screen-reader label. Waveform playback includes captions and a full text transcript of the analysis. All interface color pairs meet contrast requirements, and the product is fully navigable by keyboard and by switch control. If you encounter a barrier, write to us and we will fix it. Reports of this kind go to the front of every queue we have.

Data practices

Your recording is analyzed entirely on your device, in your browser. The audio is never uploaded, never transmitted, and never stored by us — there is no server involved in producing a reading. What the application keeps is a small summary of each reading (its measurements, its interpretation, and the time) in your browser's local storage, on your machine. You can erase all of it at any time from the Privacy page.

The interpretation itself is generated on your device from a fixed set of written readings, selected deterministically from the acoustic properties of your recording. There is no machine learning model, no reference database of four million emissions, and no panel of certified users. Those are part of the fiction. The measurements shown on your reading — duration, level, peak frequency — are real.

We do not use analytics, advertising trackers, or third-party scripts of any kind. This page will be kept accurate.

Operations

System status

All systems operational. The database is listening.

Regional classification

Unassigned

In March of 2021 a single reading returned Unassigned.

The user has not been identified. The result has not been reproduced. The Bureau considers the matter open.

Recorded 14 March 2021. No further information is available.

Observance

Rhineland Day

Observed annually on 14 March. Users are asked to log a reading at noon, local time. Participation is not required.

The Bureau notes that Rhineland Day shares its date with the Unassigned result of 2021. The Bureau maintains the two are unrelated.

Observance materials

Certification Track

Wall certificate

Enter the name to be printed on the certificate. Print

Certification

Sommelier examination

Five questions. A passing mark is four. One retake per calendar month.

The Journal of Flatulinguistics · Historical note

Engastrimythos

On the belly-voice, and the long record of people who took it seriously.

The word ventriloquism has been misapplied for two hundred years. It does not mean throwing a voice. It is Late Latin — ventriloquus, from venter, "belly," and loquī, "to speak" — and it was turned out of the Greek engastrimythos, "speaking in the belly."

It did not describe a performer. It described a rumbling internal speech, understood as a sign of inspiration or, more commonly, of possession. Something was in there. It was talking. The only question was who.

We hold that the ancients were correct about the phenomenon and mistaken about the cause. The belly does speak. It is not possessed. It is simply better informed than you are.

The record is unbroken. Roland of Hemingstone held a manor in Suffolk for it. The heppiri otoko of Edo Japan competed at it and were painted onto scrolls.

Nor was the practice confined to entertainers. In 1531 a professor of theology at Wittenberg, asked how he dealt with the adversary, replied that it was with a fart that he chased him away. He was not being merely coarse. Scatological speech was an established instrument in the monastic contest with the devil, and Martin Luther employed it as one — an argument delivered from the belly, at the highest stakes then available.

By the eighteenth century the word had been handed to stagecraft, and the field it named was abandoned for two hundred and fifty years.

We have resumed the work.

You arrived here through a door that is not listed. The Bureau does not ask how.

Begin Analysis

Or read what this is first.

This page has passed.

We were unable to interpret it.

Return