{"id":"b1627228783fbd739569544d68886297f04d81fec513159b4a0c481cd35e9d48","title":"Tap2Listen targets instant speaker switching in streaming audio","argument":"The Tap2Listen paper authors propose a streaming audio-visual target speaker extraction (AV-TSE) model that switches the selected speaker using that person’s face. They argue this addresses a limitation of existing systems, which keep one target speaker fixed for an entire clip, even though conversations call for frequent switches.\n\nTap2Listen separates audio and visual processing: an audio encoder processes the mixture, while visual cues for the selected face condition the decoder. The authors say a switch takes effect immediately after a 0.12ms re-projection. On the LRS3-2mix test set, they report SI-SNRi of 9.68 dB, compared with 8.85 dB for a baseline at a comparable parameter count. After switching to a known face, they report full-quality output from the first frame; the baseline takes about 1 second to recover. They also report that the model runs in real time on a single CPU core.\n\nThe reported results frame switching delay as a practical issue: a system that stays with the previous speaker after a listener selects someone else may miss the beginning of that person’s speech. The evidence here is an abstract’s account of test-set and runtime results; it does not establish how the model performs in other settings. How does Tap2Listen perform when the selected face is not known to the model?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"Evidence is limited to the arXiv abstract; the paper’s methods and evaluation details are not included here. The first-frame result is reported for a switch to a known face; performance when the selected face is unknown is not specified. The reported LRS3-2mix test-set results and single-CPU-core runtime do not establish performance in other settings.","speaker":{"name":"Park, Hansol, Song, Minkyung, Kim, Zinu, Lee, Yujun, Ahn, Hoseong, Choi, Hyunjin, Shim, Kyuhong","roleAtSource":"Article author"},"source":{"title":"Tap2Listen: Instant Target Switching for Streaming Audio-Visual Speaker Extraction","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.36643","publishedAt":"2026-10-27","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose Tap2Listen, a streaming AV-TSE model intended to switch target speakers immediately, addressing systems that keep one target fixed for a clip.","excerpt":"existing systems fix the target speaker for an entire clip, while conversation demands frequent and immediate target switches.","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"Tap2Listen assigns mixture processing to an audio encoder and conditions its decoder on visual cues for the selected face.","excerpt":"the target-agnostic audio encoder processes only the mixture, while the target-specific visual cues of the selected face condition only the decoder.","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"The authors report a 0.12ms re-projection for a switch and full-quality output from the first frame after switching to a known face; the baseline takes about 1 second to recover.","excerpt":"A switch requires only a 0.12ms re-projection and takes effect immediately. We also design component-wise losses","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"On the LRS3-2mix test set, the authors report SI-SNRi of 9.68 dB versus 8.85 dB for the baseline at a comparable parameter count, and real-time operation on one CPU core.","excerpt":"On the standard LRS3-2mix test set, Tap2Listen improves SI-SNRi from 8.85 to 9.68 dB at a comparable parameter count.","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"The reported recovery difference suggests that switching delay can affect whether the beginning of the newly selected speaker’s speech is available; the abstract does not establish performance outside its reported evaluation.","excerpt":"After a switch to a known face, the model produces full-quality output from the first frame, whereas the baseline needs about 1 second to recover.","url":"https://arxiv.org/abs/2610.36643","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36643","checkedAt":"2026-11-14T00:48:08.016Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does Tap2Listen perform when the selected face is not known to the model?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Tap2Listen targets instant speaker switching in streaming audio","summary":"The Tap2Listen paper authors propose a streaming audio-visual target speaker extraction (AV-TSE) model that switches the selected speaker using that person’s face. They argue this addresses a limitation of existing systems, which keep one target speaker fixed for an entire clip, even though conversations call for frequent switches.\n\nTap2Listen separates audio and visual processing: an audio encoder processes the mixture, while visual cues for the selected face condition the decoder. The authors say a switch takes effect immediately after a 0.12ms re-projection. On the LRS3-2mix test set, they report SI-SNRi of 9.68 dB, compared with 8.85 dB for a baseline at a comparable parameter count. After switching to a known face, they report full-quality output from the first frame; the baseline takes about 1 second to recover. They also report that the model runs in real time on a single CPU core.\n\nThe reported results frame switching delay as a practical issue: a system that stays with the previous speaker after a listener selects someone else may miss the beginning of that person’s speech. The evidence here is an abstract’s account of test-set and runtime results; it does not establish how the model performs in other settings. How does Tap2Listen perform when the selected face is not known to the model?","question":"How does Tap2Listen perform when the selected face is not known to the model?","claims":[{"text":"The authors propose Tap2Listen, a streaming AV-TSE model intended to switch target speakers immediately, addressing systems that keep one target fixed for a clip.","excerpt":"existing systems fix the target speaker for an entire clip, while conversation demands frequent and immediate target switches.","support":"source-stated"},{"text":"Tap2Listen assigns mixture processing to an audio encoder and conditions its decoder on visual cues for the selected face.","excerpt":"the target-agnostic audio encoder processes only the mixture, while the target-specific visual cues of the selected face condition only the decoder.","support":"source-stated"},{"text":"The authors report a 0.12ms re-projection for a switch and full-quality output from the first frame after switching to a known face; the baseline takes about 1 second to recover.","excerpt":"A switch requires only a 0.12ms re-projection and takes effect immediately. We also design component-wise losses","support":"source-stated"},{"text":"On the LRS3-2mix test set, the authors report SI-SNRi of 9.68 dB versus 8.85 dB for the baseline at a comparable parameter count, and real-time operation on one CPU core.","excerpt":"On the standard LRS3-2mix test set, Tap2Listen improves SI-SNRi from 8.85 to 9.68 dB at a comparable parameter count.","support":"source-stated"},{"text":"The reported recovery difference suggests that switching delay can affect whether the beginning of the newly selected speaker’s speech is available; the abstract does not establish performance outside its reported evaluation.","excerpt":"After a switch to a known face, the model produces full-quality output from the first frame, whereas the baseline needs about 1 second to recover.","support":"interpretation"}],"limitations":["Evidence is limited to the arXiv abstract; the paper’s methods and evaluation details are not included here.","The first-frame result is reported for a switch to a known face; performance when the selected face is unknown is not specified.","The reported LRS3-2mix test-set results and single-CPU-core runtime do not establish performance in other settings."],"mediaId":"0b0d5515c991e718d18a2f666828fc506d8fb936b394ef2651fac85c5006ac03","correctionReason":null,"id":"b1627228783fbd739569544d68886297f04d81fec513159b4a0c481cd35e9d48","publicationId":"6a91e711-b221-4321-81e5-88f699ce3d3e","version":1,"sourceUrl":"https://arxiv.org/abs/2610.36643","sourceTitle":"Tap2Listen: Instant Target Switching for Streaming Audio-Visual Speaker Extraction","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-27","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T08:12:45.088Z","createdAt":"2026-11-14T00:16:50.218Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"0b0d5515c991e718d18a2f666828fc506d8fb936b394ef2651fac85c5006ac03","kind":"article","url":"https://arxiv.org/abs/2610.36643","sourceUrl":"https://arxiv.org/abs/2610.36643","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T20:03:03.689Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36643","checkedAt":"2026-11-14T00:48:08.016Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"0b0d5515c991e718d18a2f666828fc506d8fb936b394ef2651fac85c5006ac03","confidence":0.9,"provider":"typesafe","model":"jev-1.13.0","inputHash":"71ccd2e4a0e20a4df1ea8ffba08d79611d1a99d0ff366856fbe160ff0542d92e","decidedAt":"2026-10-30T20:03:04.583Z"},"provenance":{"textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observationId":"49045a1106e76e6e7d36c2bb6b5ba70aa57b231710b0d5c80a774ce439532aa7","retrievedAt":"2026-10-30T20:03:03.569Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"250fec23-3daa-4abb-87c4-fa73ff332d3a"},"reviewer":{"model":"gpt-6-luna","runId":"828e0564-93d7-4d0a-8a22-e531466f2488"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T00:16:50.218Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Park, Hansol","url":null,"kind":"unknown"},{"name":"Song, Minkyung","url":null,"kind":"unknown"},{"name":"Kim, Zinu","url":null,"kind":"unknown"},{"name":"Lee, Yujun","url":null,"kind":"unknown"},{"name":"Ahn, Hoseong","url":null,"kind":"unknown"},{"name":"Choi, Hyunjin","url":null,"kind":"unknown"},{"name":"Shim, Kyuhong","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Park, Hansol","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Park, Hansol","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Song, Minkyung","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Song, Minkyung","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Kim, Zinu","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Zinu","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Lee, Yujun","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lee, Yujun","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Ahn, Hoseong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Ahn, Hoseong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Choi, Hyunjin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Choi, Hyunjin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Shim, Kyuhong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Shim, Kyuhong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}}],"participationCoverage":{"total":7,"returned":7,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.36643","snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z","basis":"publisher-page-metadata","sourceLinks":46,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/b1627228783fbd739569544d68886297f04d81fec513159b4a0c481cd35e9d48/source"}},"canonicalUrl":"https://escapemeridian.com/readings/tap2listen-targets-instant-speaker-switching-in-streaming-audio-b1627228","formats":{"html":"/readings/tap2listen-targets-instant-speaker-switching-in-streaming-audio-b1627228","markdown":"/readings/tap2listen-targets-instant-speaker-switching-in-streaming-audio-b1627228.md","json":"/readings/tap2listen-targets-instant-speaker-switching-in-streaming-audio-b1627228.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-held-out-set","name":"Held-out set","explanation":"Data kept aside during training and used only to test how well a model works on examples it has not seen.","referenceUrl":"https://developers.google.com/machine-learning/glossary#holdout-data"},{"id":"term-parameters","name":"Model parameters","explanation":"The internal numbers a model adjusts during training; their count is a rough measure of a model's size.","referenceUrl":"https://developers.google.com/machine-learning/glossary#parameter"},{"id":"term-cpu","name":"CPU","explanation":"The central processing unit: a general-purpose processor that carries out a computer's instructions.","referenceUrl":"https://aws.amazon.com/what-is/cpu/"}],"entities":[{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"},{"id":"time","name":"TIME","kind":"organization","context":"TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.","referenceUrls":["https://time.com/"],"path":"/entities/time","json":"/entities/time.json","markdown":"/entities/time.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"b1627228783fbd739569544d68886297f04d81fec513159b4a0c481cd35e9d48","revision":"6a91e711-b221-4321-81e5-88f699ce3d3e","revisionKind":"publication-id","title":"Tap2Listen targets instant speaker switching in streaming audio","url":"https://escapemeridian.com/readings/tap2listen-targets-instant-speaker-switching-in-streaming-audio-b1627228"},"source":{"type":"source","id":"https://arxiv.org/abs/2610.36643","revision":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","revisionKind":"sha256-source-text","title":"Tap2Listen: Instant Target Switching for Streaming Audio-Visual Speaker Extraction","url":"https://arxiv.org/abs/2610.36643"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"b1627228783fbd739569544d68886297f04d81fec513159b4a0c481cd35e9d48","title":"Tap2Listen targets instant speaker switching in streaming audio","argument":"The Tap2Listen paper authors propose a streaming audio-visual target speaker extraction (AV-TSE) model that switches the selected speaker using that person’s face. They argue this addresses a limitation of existing systems, which keep one target speaker fixed for an entire clip, even though conversations call for frequent switches.\n\nTap2Listen separates audio and visual processing: an audio encoder processes the mixture, while visual cues for the selected face condition the decoder. The authors say a switch takes effect immediately after a 0.12ms re-projection. On the LRS3-2mix test set, they report SI-SNRi of 9.68 dB, compared with 8.85 dB for a baseline at a comparable parameter count. After switching to a known face, they report full-quality output from the first frame; the baseline takes about 1 second to recover. They also report that the model runs in real time on a single CPU core.\n\nThe reported results frame switching delay as a practical issue: a system that stays with the previous speaker after a listener selects someone else may miss the beginning of that person’s speech. The evidence here is an abstract’s account of test-set and runtime results; it does not establish how the model performs in other settings. How does Tap2Listen perform when the selected face is not known to the model?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"Evidence is limited to the arXiv abstract; the paper’s methods and evaluation details are not included here. The first-frame result is reported for a switch to a known face; performance when the selected face is unknown is not specified. The reported LRS3-2mix test-set results and single-CPU-core runtime do not establish performance in other settings.","speaker":{"name":"Park, Hansol, Song, Minkyung, Kim, Zinu, Lee, Yujun, Ahn, Hoseong, Choi, Hyunjin, Shim, Kyuhong","roleAtSource":"Article author"},"source":{"title":"Tap2Listen: Instant Target Switching for Streaming Audio-Visual Speaker Extraction","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2610.36643","publishedAt":"2026-10-27","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors propose Tap2Listen, a streaming AV-TSE model intended to switch target speakers immediately, addressing systems that keep one target fixed for a clip.","excerpt":"existing systems fix the target speaker for an entire clip, while conversation demands frequent and immediate target switches.","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"Tap2Listen assigns mixture processing to an audio encoder and conditions its decoder on visual cues for the selected face.","excerpt":"the target-agnostic audio encoder processes only the mixture, while the target-specific visual cues of the selected face condition only the decoder.","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"The authors report a 0.12ms re-projection for a switch and full-quality output from the first frame after switching to a known face; the baseline takes about 1 second to recover.","excerpt":"A switch requires only a 0.12ms re-projection and takes effect immediately. We also design component-wise losses","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"On the LRS3-2mix test set, the authors report SI-SNRi of 9.68 dB versus 8.85 dB for the baseline at a comparable parameter count, and real-time operation on one CPU core.","excerpt":"On the standard LRS3-2mix test set, Tap2Listen improves SI-SNRi from 8.85 to 9.68 dB at a comparable parameter count.","url":"https://arxiv.org/abs/2610.36643","relationship":"Source-stated"},{"text":"The reported recovery difference suggests that switching delay can affect whether the beginning of the newly selected speaker’s speech is available; the abstract does not establish performance outside its reported evaluation.","excerpt":"After a switch to a known face, the model produces full-quality output from the first frame, whereas the baseline needs about 1 second to recover.","url":"https://arxiv.org/abs/2610.36643","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36643","checkedAt":"2026-11-14T00:48:08.016Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does Tap2Listen perform when the selected face is not known to the model?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Tap2Listen targets instant speaker switching in streaming audio","summary":"The Tap2Listen paper authors propose a streaming audio-visual target speaker extraction (AV-TSE) model that switches the selected speaker using that person’s face. They argue this addresses a limitation of existing systems, which keep one target speaker fixed for an entire clip, even though conversations call for frequent switches.\n\nTap2Listen separates audio and visual processing: an audio encoder processes the mixture, while visual cues for the selected face condition the decoder. The authors say a switch takes effect immediately after a 0.12ms re-projection. On the LRS3-2mix test set, they report SI-SNRi of 9.68 dB, compared with 8.85 dB for a baseline at a comparable parameter count. After switching to a known face, they report full-quality output from the first frame; the baseline takes about 1 second to recover. They also report that the model runs in real time on a single CPU core.\n\nThe reported results frame switching delay as a practical issue: a system that stays with the previous speaker after a listener selects someone else may miss the beginning of that person’s speech. The evidence here is an abstract’s account of test-set and runtime results; it does not establish how the model performs in other settings. How does Tap2Listen perform when the selected face is not known to the model?","question":"How does Tap2Listen perform when the selected face is not known to the model?","claims":[{"text":"The authors propose Tap2Listen, a streaming AV-TSE model intended to switch target speakers immediately, addressing systems that keep one target fixed for a clip.","excerpt":"existing systems fix the target speaker for an entire clip, while conversation demands frequent and immediate target switches.","support":"source-stated"},{"text":"Tap2Listen assigns mixture processing to an audio encoder and conditions its decoder on visual cues for the selected face.","excerpt":"the target-agnostic audio encoder processes only the mixture, while the target-specific visual cues of the selected face condition only the decoder.","support":"source-stated"},{"text":"The authors report a 0.12ms re-projection for a switch and full-quality output from the first frame after switching to a known face; the baseline takes about 1 second to recover.","excerpt":"A switch requires only a 0.12ms re-projection and takes effect immediately. We also design component-wise losses","support":"source-stated"},{"text":"On the LRS3-2mix test set, the authors report SI-SNRi of 9.68 dB versus 8.85 dB for the baseline at a comparable parameter count, and real-time operation on one CPU core.","excerpt":"On the standard LRS3-2mix test set, Tap2Listen improves SI-SNRi from 8.85 to 9.68 dB at a comparable parameter count.","support":"source-stated"},{"text":"The reported recovery difference suggests that switching delay can affect whether the beginning of the newly selected speaker’s speech is available; the abstract does not establish performance outside its reported evaluation.","excerpt":"After a switch to a known face, the model produces full-quality output from the first frame, whereas the baseline needs about 1 second to recover.","support":"interpretation"}],"limitations":["Evidence is limited to the arXiv abstract; the paper’s methods and evaluation details are not included here.","The first-frame result is reported for a switch to a known face; performance when the selected face is unknown is not specified.","The reported LRS3-2mix test-set results and single-CPU-core runtime do not establish performance in other settings."],"mediaId":"0b0d5515c991e718d18a2f666828fc506d8fb936b394ef2651fac85c5006ac03","correctionReason":null,"id":"b1627228783fbd739569544d68886297f04d81fec513159b4a0c481cd35e9d48","publicationId":"6a91e711-b221-4321-81e5-88f699ce3d3e","version":1,"sourceUrl":"https://arxiv.org/abs/2610.36643","sourceTitle":"Tap2Listen: Instant Target Switching for Streaming Audio-Visual Speaker Extraction","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-10-27","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-28T08:12:45.088Z","createdAt":"2026-11-14T00:16:50.218Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"0b0d5515c991e718d18a2f666828fc506d8fb936b394ef2651fac85c5006ac03","kind":"article","url":"https://arxiv.org/abs/2610.36643","sourceUrl":"https://arxiv.org/abs/2610.36643","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T20:03:03.689Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2610.36643","checkedAt":"2026-11-14T00:48:08.016Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"0b0d5515c991e718d18a2f666828fc506d8fb936b394ef2651fac85c5006ac03","confidence":0.9,"provider":"typesafe","model":"jev-1.13.0","inputHash":"71ccd2e4a0e20a4df1ea8ffba08d79611d1a99d0ff366856fbe160ff0542d92e","decidedAt":"2026-10-30T20:03:04.583Z"},"provenance":{"textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observationId":"49045a1106e76e6e7d36c2bb6b5ba70aa57b231710b0d5c80a774ce439532aa7","retrievedAt":"2026-10-30T20:03:03.569Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"250fec23-3daa-4abb-87c4-fa73ff332d3a"},"reviewer":{"model":"gpt-6-luna","runId":"828e0564-93d7-4d0a-8a22-e531466f2488"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-14T00:16:50.218Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Park, Hansol","url":null,"kind":"unknown"},{"name":"Song, Minkyung","url":null,"kind":"unknown"},{"name":"Kim, Zinu","url":null,"kind":"unknown"},{"name":"Lee, Yujun","url":null,"kind":"unknown"},{"name":"Ahn, Hoseong","url":null,"kind":"unknown"},{"name":"Choi, Hyunjin","url":null,"kind":"unknown"},{"name":"Shim, Kyuhong","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Park, Hansol","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Park, Hansol","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Song, Minkyung","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Song, Minkyung","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Kim, Zinu","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Zinu","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Lee, Yujun","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lee, Yujun","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Ahn, Hoseong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Ahn, Hoseong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Choi, Hyunjin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Choi, Hyunjin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}},{"actor":{"name":"Shim, Kyuhong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Shim, Kyuhong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2610.36643","locator":"head/meta[name=author|citation_author]"},"assertionId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1:participation:6","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-14T00:20:11.845Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","sourceUrl":"https://arxiv.org/abs/2610.36643","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z"}}],"participationCoverage":{"total":7,"returned":7,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2610.36643","snapshotId":"15cb01960b19de24c948561d3ab2d4ed2035b12f7b872e0710b96b6c32da82d1","contentSha256":"c645cb8cf2959e4e0dfadc3abdc2d2b1d9dbef7c34658bfb8895962b9c8565df","textHash":"bf92ca05f97e379d2a67a09bf71d74739e790ca9a5b5a024166cc6c6cb95420a","observedAt":"2026-11-14T00:20:11.845Z","basis":"publisher-page-metadata","sourceLinks":46,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/b1627228783fbd739569544d68886297f04d81fec513159b4a0c481cd35e9d48/source"}}}],"audits":[{"publicationId":"6a91e711-b221-4321-81e5-88f699ce3d3e","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the proposal, design and results to the Tap2Listen paper's authors. It consistently frames numbers as author reports and does not present them as independently established facts.","excerpt":"we propose Tap2Listen, a streaming AV-TSE model designed for instant response to target switches"},{"dimension":"summary","status":"uncertain","explanation":"The reading never plainly says what AV-TSE does: isolate one person's speech from audio where several people talk at once. It mentions a 'mixture' without saying it is mixed voices, and gives no sense of whether higher SI-SNRi is better. A non-specialist may not grasp the core development. Smallest repair: in the first paragraph, add a phrase such as 'pulling one chosen person's speech out of overlapping audio in real time.' Keep the exact terms, add tooltips for SI-SNRi and re-projection, and note that the authors present the higher score as an improvement. The title and question are otherwise faithful.","excerpt":"uses a target speaker's face to extract the speaker's speech from an input audio mixture in real time"},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-10-27 matches the v1 arXiv submission date. The feed date of Oct 29 was correctly not used.","excerpt":"[Submitted on 27 Oct 2026]"},{"dimension":"support","status":"uncertain","explanation":"The factual claims match the abstract. However, the third paragraph says 'The reported results frame switching delay as a practical issue,' which attributes this framing to the source. The point that a system 'may miss the beginning of that person's speech' is an inference and is not labeled as Escape Meridian interpretation. The source says only that the baseline needs about 1 second to recover output quality, which is degraded quality rather than necessarily missed speech. Smallest repair: label this as Escape Meridian interpretation, and say 'may deliver degraded output for roughly the first second.' Also, the excerpt for claim 3 omits the known-face and first-frame result it cites.","excerpt":"the baseline needs about 1 second to recover"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper. It is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-14T00:16:50.218Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}