https://github.com/facebookresearch/audioseal

you can also pip install audioseal

\n","updatedAt":"2024-01-31T09:22:54.587Z","author":{"_id":"60338cdfb5883695ce9d0916","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60338cdfb5883695ce9d0916/BXbJKt_UwoFnh7xB4ewHs.jpeg","fullname":"Hady Elsahar","name":"hadyelsahar","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":21}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7999864816665649},"editors":["hadyelsahar"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/60338cdfb5883695ce9d0916/BXbJKt_UwoFnh7xB4ewHs.jpeg"],"reactions":[],"isReport":false}},{"id":"65ba317798bd7a573be377eb","author":{"_id":"644e1b1d9b4e87c31bab0a14","avatarUrl":"/avatars/88bb4c4a67dc8958069e9014f5e73a0b.svg","fullname":"Michael Barry","name":"MichaelBarryUK","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":7},"createdAt":"2024-01-31T11:39:35.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Are the troublemakers compelled to use watermarked LLMs, or are they free to use non-watermarked models? ","html":"

Are the troublemakers compelled to use watermarked LLMs, or are they free to use non-watermarked models?

\n","updatedAt":"2024-01-31T11:39:35.420Z","author":{"_id":"644e1b1d9b4e87c31bab0a14","avatarUrl":"/avatars/88bb4c4a67dc8958069e9014f5e73a0b.svg","fullname":"Michael Barry","name":"MichaelBarryUK","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":7}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.903719425201416},"editors":["MichaelBarryUK"],"editorAvatarUrls":["/avatars/88bb4c4a67dc8958069e9014f5e73a0b.svg"],"reactions":[],"isReport":false}},{"id":"65ba34c01d7d1717cc4dd435","author":{"_id":"6032802e1f993496bc14d9e3","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png","fullname":"Omar Sanseviero","name":"osanseviero","type":"user","isPro":false,"isHf":false,"isHfAdmin":true,"isMod":false,"followerCount":3289},"createdAt":"2024-01-31T11:53:36.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Interesting! Check this out @reach-vb ","html":"

Interesting! Check this out \n\n@reach-vb\n\t

\n","updatedAt":"2024-01-31T11:53:36.527Z","author":{"_id":"6032802e1f993496bc14d9e3","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png","fullname":"Omar Sanseviero","name":"osanseviero","type":"user","isPro":false,"isHf":false,"isHfAdmin":true,"isMod":false,"followerCount":3289}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.46625614166259766},"editors":["osanseviero"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6032802e1f993496bc14d9e3/w6hr-DEQot4VVkoyRIBiy.png"],"reactions":[],"isReport":false}},{"id":"65baf2e5b30c170eb1f32376","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":264},"createdAt":"2024-02-01T01:24:53.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [Detecting Voice Cloning Attacks via Timbre Watermarking](https://huggingface.co/papers/2312.03410) (2023)\n* [Benchmarking the Robustness of Image Watermarks](https://huggingface.co/papers/2401.08573) (2024)\n* [Cross-Attention Watermarking of Large Language Models](https://huggingface.co/papers/2401.06829) (2024)\n* [Robust Image Watermarking using Stable Diffusion](https://huggingface.co/papers/2401.04247) (2024)\n* [EditGuard: Versatile Image Watermarking for Tamper Localization and Copyright Protection](https://huggingface.co/papers/2312.08883) (2023)\n\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space\n\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: `@librarian-bot recommend`","html":"

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Detecting Voice Cloning Attacks via Timbre Watermarking (2023)
Benchmarking the Robustness of Image Watermarks (2024)
Cross-Attention Watermarking of Large Language Models (2024)
Robust Image Watermarking using Stable Diffusion (2024)
EditGuard: Versatile Image Watermarking for Tamper Localization and Copyright Protection (2023)

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: \n\n@librarian-bot\n\t recommend

\n","updatedAt":"2024-02-01T01:24:53.910Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":264}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7216165661811829},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2401.17264","authors":[{"_id":"65b9cf6dd88a65abb9117e6a","user":{"_id":"65ba260b3167e14767093ca7","avatarUrl":"/avatars/f23fa3e8429ac603dd512cc52481a9bc.svg","isPro":false,"fullname":"Robin San Roman","user":"Sparker17","type":"user"},"name":"Robin San Roman","status":"claimed_verified","statusLastChangedAt":"2024-01-31T11:48:18.242Z","hidden":false},{"_id":"65b9cf6dd88a65abb9117e6b","user":{"_id":"616c22e040e2f69baa1c7ae0","avatarUrl":"/avatars/d6c00e7d98e5e8a52e99aa7b1a7815ee.svg","isPro":false,"fullname":"Pierre Fernandez","user":"pierrefdz","type":"user"},"name":"Pierre Fernandez","status":"admin_assigned","statusLastChangedAt":"2024-01-31T10:03:32.247Z","hidden":false},{"_id":"65b9cf6dd88a65abb9117e6c","user":{"_id":"6357f5f50dce5559211a3503","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1666708948380-noauth.jpeg","isPro":false,"fullname":"Alexandre Défossez","user":"adefossez","type":"user"},"name":"Alexandre Défossez","status":"admin_assigned","statusLastChangedAt":"2024-01-31T10:03:41.720Z","hidden":false},{"_id":"65b9cf6dd88a65abb9117e6d","user":{"_id":"670000393ec4cc293d41d92c","avatarUrl":"/avatars/b67843e1f7addaf6769c2cba868bf501.svg","isPro":false,"fullname":"Teddy Furon","user":"tfuron","type":"user"},"name":"Teddy Furon","status":"claimed_verified","statusLastChangedAt":"2024-10-04T20:12:02.474Z","hidden":false},{"_id":"65b9cf6dd88a65abb9117e6e","user":{"_id":"6417f443058f65de4320716e","avatarUrl":"/avatars/3377a2f5431ff7f5e9d106bc5eec14e8.svg","isPro":false,"fullname":"Tuan Tran","user":"tuantran","type":"user"},"name":"Tuan Tran","status":"admin_assigned","statusLastChangedAt":"2024-01-31T10:04:18.019Z","hidden":false},{"_id":"65b9cf6dd88a65abb9117e6f","user":{"_id":"60338cdfb5883695ce9d0916","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60338cdfb5883695ce9d0916/BXbJKt_UwoFnh7xB4ewHs.jpeg","isPro":false,"fullname":"Hady Elsahar","user":"hadyelsahar","type":"user"},"name":"Hady Elsahar","status":"admin_assigned","statusLastChangedAt":"2024-01-31T10:04:24.497Z","hidden":false}],"publishedAt":"2024-01-30T18:56:22.000Z","submittedOnDailyAt":"2024-01-31T02:11:17.940Z","title":"Proactive Detection of Voice Cloning with Localized Watermarking","submittedOnDailyBy":{"_id":"60f1abe7544c2adfd699860c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674929746905-60f1abe7544c2adfd699860c.jpeg","isPro":false,"fullname":"AK","user":"akhaliq","type":"user"},"summary":"In the rapidly evolving field of speech generative models, there is a\npressing need to ensure audio authenticity against the risks of voice cloning.\nWe present AudioSeal, the first audio watermarking technique designed\nspecifically for localized detection of AI-generated speech. AudioSeal employs\na generator/detector architecture trained jointly with a localization loss to\nenable localized watermark detection up to the sample level, and a novel\nperceptual loss inspired by auditory masking, that enables AudioSeal to achieve\nbetter imperceptibility. AudioSeal achieves state-of-the-art performance in\nterms of robustness to real life audio manipulations and imperceptibility based\non automatic and human evaluation metrics. Additionally, AudioSeal is designed\nwith a fast, single-pass detector, that significantly surpasses existing models\nin speed - achieving detection up to two orders of magnitude faster, making it\nideal for large-scale and real-time applications.","upvotes":18,"discussionId":"65b9cf6dd88a65abb9117e7f","ai_summary":"AudioSeal is an audio watermarking technique that uses a generator/detector architecture to detect AI-generated speech with high robustness and imperceptibility, offering fast detection.","ai_keywords":["generator/detector architecture","localization loss","perceptual loss","auditory masking","state-of-the-art","robustness","imperceptibility","automatic evaluation","human evaluation","single-pass detector","real-time applications"]},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"611997b74896e2eb1dc3db85","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/611997b74896e2eb1dc3db85/Rtm_270HBxnLm-cKmV4J6.jpeg","isPro":false,"fullname":"Valério Cardoso","user":"valeriocardoso","type":"user"},{"_id":"63f60822b607296857b44e03","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63f60822b607296857b44e03/7861Kb3BPs9ctltqyOAzG.jpeg","isPro":false,"fullname":"Tom Osman","user":"tomosman","type":"user"},{"_id":"60338cdfb5883695ce9d0916","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60338cdfb5883695ce9d0916/BXbJKt_UwoFnh7xB4ewHs.jpeg","isPro":false,"fullname":"Hady Elsahar","user":"hadyelsahar","type":"user"},{"_id":"616c22e040e2f69baa1c7ae0","avatarUrl":"/avatars/d6c00e7d98e5e8a52e99aa7b1a7815ee.svg","isPro":false,"fullname":"Pierre Fernandez","user":"pierrefdz","type":"user"},{"_id":"65ba238421afe0ec206952ca","avatarUrl":"/avatars/0f2b6ccfd6dce69bfe03853819fb5903.svg","isPro":false,"fullname":"Tuan Tran","user":"antoinetran","type":"user"},{"_id":"65ba260b3167e14767093ca7","avatarUrl":"/avatars/f23fa3e8429ac603dd512cc52481a9bc.svg","isPro":false,"fullname":"Robin San Roman","user":"Sparker17","type":"user"},{"_id":"635964636a61954080850e1d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/635964636a61954080850e1d/0bfExuDTrHTtm8c-40cDM.png","isPro":false,"fullname":"William Lamkin","user":"phanes","type":"user"},{"_id":"5e1f1a7d5c2e2c73f4512186","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/5e1f1a7d5c2e2c73f4512186/yYe60T8GOjGjKSJRpY8vb.jpeg","isPro":false,"fullname":"Zhanliang Liu","user":"zliu","type":"user"},{"_id":"6340e58d79f2908105fbfa92","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6340e58d79f2908105fbfa92/ByAhqdxDqhHjHBcerPZTx.jpeg","isPro":false,"fullname":"Anthony Guijarro","user":"Anthonyg5005","type":"user"},{"_id":"6538119803519fddb4a17e10","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6538119803519fddb4a17e10/ffJMkdx-rM7VvLTCM6ri_.jpeg","isPro":false,"fullname":"samusenps","user":"samusenps","type":"user"},{"_id":"6366313c361a96184dbadff8","avatarUrl":"/avatars/9b83c5aedc02267d9596b19c20fbe593.svg","isPro":false,"fullname":"HAN JUNGU","user":"JUNGU","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0}">

Papers

arxiv:2401.17264

Proactive Detection of Voice Cloning with Localized Watermarking

Published on Jan 30, 2024

· Submitted by

AK on Jan 31, 2024

Upvote

Authors:

Robin San Roman ,

Pierre Fernandez ,

Alexandre Défossez ,

Teddy Furon ,

Tuan Tran ,

Hady Elsahar

Abstract

AudioSeal is an audio watermarking technique that uses a generator/detector architecture to detect AI-generated speech with high robustness and imperceptibility, offering fast detection.

AI-generated summary

In the rapidly evolving field of speech generative models, there is a pressing need to ensure audio authenticity against the risks of voice cloning. We present AudioSeal, the first audio watermarking technique designed specifically for localized detection of AI-generated speech. AudioSeal employs a generator/detector architecture trained jointly with a localization loss to enable localized watermark detection up to the sample level, and a novel perceptual loss inspired by auditory masking, that enables AudioSeal to achieve better imperceptibility. AudioSeal achieves state-of-the-art performance in terms of robustness to real life audio manipulations and imperceptibility based on automatic and human evaluation metrics. Additionally, AudioSeal is designed with a fast, single-pass detector, that significantly surpasses existing models in speed - achieving detection up to two orders of magnitude faster, making it ideal for large-scale and real-time applications.