Frequently asked questions
General
What does VERA™ verify?
Whether a piece of music was produced by a specific AI generator. You get more than an AI-or-not answer. Every generator builds sound in its own recognizable way, what we call its signature, so a result tells you which one made the track and shows you the measurement behind it. Suno is supported in production today. More generators are on the way, and they come with your plan at no extra cost.
Which versions of Suno does VERA™ cover?
Every Suno version released so far. A result names the generator, so a match tells you it was Suno rather than just "AI". More generators are on the way, and they come with your plan at no extra cost.
What's next for VERA™?
A lot, and some of it is big. VERA runs on research we do ourselves, and that research is moving fast. More generators are on the way, and each one comes with your plan at no extra cost. Beyond that, we're building things nobody in this field can do yet. Soon a result won't just name the generator. It will show you how that generator relates to the others. It's the most exciting work we've done, and we'll share it as soon as it ships.
What is an AI music detector, and how is VERA™ different?
An AI music detector runs a machine-learning model over a recording and estimates whether a generative service made it. You get a probability for the track as a whole. We start by taking the track apart instead. Each stem is read on its own for the traces a particular generator leaves in the signal, so a generated part buried in an otherwise human-made mix still shows up rather than averaging out across the mix. And the result is deterministic. It tells you which generator made the track and brings the measurements with it.
Is VERA™ an AI music detector?
Not in the usual sense. As far as we know, every AI music detector on the market today is a classifier. They estimate a probability, usually from a model trained on a lot of audio, and usually across the whole mix at once. VERA is a verification tool, built from measurement methods we made ourselves. We separate the stems, check each one for the known pattern a specific generator leaves behind, and plot what we found against the reference for that generator. There is a number in the report as well, but it measures how closely your track matches that reference, not how likely it is to be AI.
What is the best AI music detector?
That depends on what you will do with the answer. If you need a quick read on a lot of audio, so a person knows where to look first, a detector that gives a score is built for that. If the answer decides a payout or an appeal, you need a result you can show someone. Reliable means something different for each. A score is reliable if it is usually right. A finding is reliable if it still stands when someone questions it. A few questions help you choose. Does it name the generator, or only say "AI"? Does it check the vocals, the drums and the rest of the mix one at a time? What happens when a generator is updated? And when an artist asks how you know, what do you open? VERA™, by Oddtune, is built for the second kind of decision. It is a verification tool rather than a detector. It names the generator and shows its pattern in your track, drawn over that generator's reference. Many teams run both.
Can VERA™ distinguish fully AI-generated tracks from partially AI-generated ones?
Yes. A track-level result covers the fully generated case, and stem separation covers the partial ones. We split the recording into stem groups (vocals, drums or other) and trace the evidence back to whichever one carries it, so you can see which side of the mix the AI is on. Vocals over human production is the obvious case, and it is also the one a whole-mix tool is most likely to let through. One generated part inside an otherwise human-made mix gets diluted by everything around it. Separating first keeps it legible. We validated the separation method for this purpose, because it has to be calibrated to the measurement for the evidence to hold. Broader AI-assisted workflows, such as AI-aided composition rendered through human production, sit outside what we report today.
Is analysis done on the full track or on segments?
The complete recording. You get a single result per supported generator with its evidence, not a result from an isolated slice. Stem separation is a different cut of the same analysis, splitting the recording into stem groups rather than into time segments.
How accurate is VERA™?
Accuracy percentages are a machine-learning measure of how often a model's predictions matched a labeled test set. We measure the recording rather than predicting from it, so the question is different. Is a known generator pattern present in this audio, and can you see it? Every result comes with the measured signature plotted against the reference it was compared with. The figure we watch hardest is how often authentic music gets flagged, because a wrong flag lands on a real creator. Across a large reference set of human-made music, fewer than 0.04% of tracks came back flagged. Read that as a ceiling, not a rate. In customer use to date there are zero confirmed false positives.
How it works
What does VERA™ actually measure in a track?
The structure of the audio itself. AI generators build sound on a fixed grid of tiny, equal steps. That grid leaves an evenly spaced pattern in the signal, the same in every song a generator makes. Human recordings are naturally uneven and carry no such pattern. We read it straight from the file, whatever the genre or style.
Why doesn't VERA™ use AI to identify AI music?
Because a trained classifier can only estimate. It learns what AI music tends to sound like from a dataset, then scores new tracks against that impression. The score can be right and still leave you nothing to show, and when a generator is retrained the classifier has to be retrained too. We measure the recording against a signature that comes from the generator's architecture, not its training data.
What is the difference between "probabilistic" and "deterministic" verification?
A probabilistic tool learns patterns from examples and reports how likely a track is to be AI. We are deterministic. The same fixed measurements run on every recording, and what they found is in the result, not just the figure that sums it up.
How is evidence from VERA™ different from a confidence score?
A confidence score summarizes a model's estimate in a single number. Our evidence is a physical measurement of the recording, the spectral signature left by the generator's architecture, plotted against the known reference for that generator.
Does VERA™ work if the AI model is updated or fine-tuned?
Yes. The signatures we measure come from a generator's core architecture, not its trained weights. Fine-tuning or retraining on new data changes the weights, not the architecture, so the traces stay put and the evidence holds.
Using VERA™
How can I check if a song is AI-generated?
Upload the track to VERA™ and run an analysis. You get a finding on whether a supported AI generator made the song, which generator produced it, and a report with the evidence attached. Free AI song detectors read the whole mix and give you a probability. We take the track apart into stems first and check each one on its own. That matters most on hybrid tracks, part human and part generated. We have had cases where even enterprise detectors found nothing suspicious and we measured the signature clearly in one or more stems.
How can I tell if a song was made with Suno?
Run it through VERA™. Suno is supported in production today. A generator like Suno builds its audio on a fixed grid, in tiny equal steps, many per second. That leaves a faint row of evenly spaced peaks in the sound, at a spacing that belongs to that generator. VERA measures the pattern straight from the file and draws it over Suno's reference, so you can see whether the peaks land on it. The vocals, the drums and the rest of the mix are checked one at a time, so a Suno vocal over a live band still shows up.
What are the input requirements?
A raw audio file supplied by you: full tracks only (no clips or snippets), a minimum length of 1 minute, and a minimum bitrate of 128 kbps, in MP3, WAV, AAC, AIFF, or FLAC. Those limits exist so there is enough signal in the file to measure. It does not matter where the track came from, and metadata is not required, since the evidence is read out of the audio itself. The API accepts exactly the same input as the dashboard.
What evidence comes with each result?
Every analysis produces a report. At its heart is the visual evidence, the signature measured in your track drawn over the reference for the generator it matched. The peaks either land on the reference or they miss it. Alongside that sit the condition checks that ran and an evidence strength, a "how close" number rather than a "how likely" one, taken from direct measurement of the audio rather than from a neural network. The report also carries the ordinary bookkeeping, track metadata, a hash of the exact file analyzed, a case ID and a timestamp, and the whole thing downloads as a PDF.
Do I need to be technical to understand VERA™ results?
No. The evidence is visual, with plain explanations next to it. Either the measured signature lines up with the reference or it does not, and spotting the difference takes no background in audio.
What does it mean if a track I believe is human-made comes back flagged?
It means the pattern we look for was measured in that recording. We do not form an impression of a track or judge whether something sounds AI-made. What a measurement cannot tell you by itself is how the pattern got there. Most of the time that is plain. When it is not, more than one honest explanation fits. The recording may carry more generated material than expected. The audio may have gone through a generator at some point, for example a remaster tool that re-generates the sound, which leaves the pattern even when people wrote and played the song. Or the pattern may have arisen for a reason we have not yet identified. We treat every case like that as ours to explain, so the report shows the measurements behind the finding, and tracks like these go into our own false-positive figure. Send us one and we will go through the evidence with you.
What does it mean if a track comes back without a match?
It means none of the signatures we measure for turned up in that file. A few things can hide one even when a generator was used. A generated part can sit low in the mix under real instruments, too quiet to read. Heavy effects like chorus or phaser smear the structure we measure, and significant pitch shifting moves it. Splitting the stems catches most of that, but something quiet enough can still sit under the floor. So when we come back with a generator, you can rely on it. When we come back with nothing, the honest word is inconclusive, not clean.
How is VERA™ priced?
Plans are sized by how many tracks you verify. You pay monthly, or for a year upfront at a lower rate, by invoice or business card. If you go past your plan's allowance, extra verifications are billed one at a time. Larger organizations get an Enterprise plan built around their volume. Fill in the Get access form and we'll get back to you with pricing.
Can we try VERA™ before committing to a bigger plan?
Yes. The smallest plan is made for that. It runs month to month with no minimum term. Take it for a month, run your own tracks through it, and decide from there. A good test is a set where you already know the answers, some human-made and some generated. You'll see what VERA finds in each, and the evidence behind it. There's no free version. The smallest plan runs the same analysis as every other plan.
How do I get started?
Tell us about your team and what you need, and we'll email you within one business day with next steps.
Developers & API
Is an API available?
Yes, and it supports high-volume processing. Keys are created, viewed, and rotated through the Developers portal in the dashboard, alongside the full documentation. API activity and dashboard activity sit under one organization view, so your team sees a single history no matter how a track was submitted.
What information does the API return?
Each analysis returns the result for the track plus supporting evidence: an evidence strength, visual evidence, and track metadata. Every analysis is also recorded in your organization's history in the dashboard, where results can be searched, filtered, organized into folders, and annotated with notes. The full response schema is in the API documentation.