ElevenLabs AI Speech Classifier

The stated limit is a duration. "We analyze the first minute of each sample" and "You can upload common audio and video files", per the product page FAQ (elevenlabs.io/ai-speech-classifier, read 10 November 2026). The 2024 version of the page printed a size cap, "Upload audio file, up to 10 MB", and "If your file is over 1 minute long, only the first minute will be analysed" (web.archive.org, 1 June 2024, read 10 November 2026); the live page prints no size and no format list, so the file types field on this card stays empty. No daily quota is published for either tool.

Nothing is sold. "The AI Speech Classifier is a free tool", says the product FAQ, and the documentation adds "the tool is free for all to use" and "the audio detector is free for all signed-in users" (elevenlabs.io/docs, Audio Detector, read 10 November 2026). The vendor's pricing page prices speech generation, not detection, so the plans table below carries the two free routes and no price. No classifier endpoint appears in the API reference or the documentation sitemap read 10 November 2026; the 2023 launch post said the vendor was "eager to collaborate with interested partners on deeper API integration", which the card records as prose, and the API field is marked unavailable.

Retention is stated for voice data in general, not for classifier uploads. The Privacy Policy (Updated 9 October 2026) says "We generally keep your personal information for as long as necessary to provide you with our Services" and "We won't keep data we generate about your voice for longer than 3 years after your last interaction with us, unless the law requires or allows otherwise" (elevenlabs.io/privacy-policy, read 10 November 2026). The product FAQ adds "We process your sample only to return a result. See our privacy policy for details on how we handle uploaded data." The three-year clause governs data the vendor derives about a voice, so the retention rank field on this card stays empty, the reading the index gave Pindrop's voice features.

The figures differ by page and by year. The documentation FAQ states "Our classifier maintains high accuracy (99% precision, 80% recall) for audio files generated with ElevenLabs that have not been modified" (elevenlabs.io/docs, Audio Detector, read 10 November 2026). The blog post "Audio deepfake detection: How it works & how to verify ElevenLabs audio" (Published Oct 14, 2026) repeats "ElevenLabs' Speech Classifier has 99% precision and 80% recall" (elevenlabs.io/blog, read 10 November 2026). The launch post of Jun 15, 2023 stated a different measure: "currently maintaining >99% (accuracy) if the input was unmodified. If it underwent Codec or reverb transformations, our Classifier maintains over 90% accuracy. This figure drops the more the content has been post-processed" (web.archive.org, 30 August 2023, read 10 November 2026). No dataset, sample count or test date accompanies any of the three.

A precision of 99% and a recall of 80% describe two different errors. Precision says how many flagged clips were ElevenLabs audio; recall says how many ElevenLabs clips were flagged, so the vendor's own figure implies about one unmodified ElevenLabs clip in five is missed. The documentation says as much in words: the classifier "is statistical and may produce false positives or miss modified audio", and the October 2026 blog post adds "Its accuracy will also drop when analyzing heavily compressed or edited speech content" (elevenlabs.io/blog, read 10 November 2026). No false positive rate is published. The claim gap field on this card stays empty because a precision is not an accuracy percentage, the treatment the index gave Binoculars.

The caveats are the vendor's own. The product page carries the banner "Does not reliably classify audio generated with the Eleven v3 model" and the FAQ states the tool "cannot detect audio from other providers" (elevenlabs.io/ai-speech-classifier, read 10 November 2026). The documentation notes that "content generated by third-party models available through ElevenLabs is generally not watermarked", so the Audio Detector's watermark check does not cover every clip made on the platform either. The safety page describes the classifier as "an industry leading AI Audio Classifier to help others determine whether a piece of content was generated using ElevenLabs" (elevenlabs.io/safety, read 10 November 2026).