{"id":232,"date":"2025-09-01T11:45:45","date_gmt":"2025-09-01T11:45:45","guid":{"rendered":"https:\/\/technewscentury.co.uk\/?p=232"},"modified":"2025-09-01T11:45:45","modified_gmt":"2025-09-01T11:45:45","slug":"honor-magic-v5-introduces-on-device-multilingual-call-translation","status":"publish","type":"post","link":"https:\/\/technewscentury.co.uk\/?p=232","title":{"rendered":"HONOR Magic V5 Introduces On-Device Multilingual Call Translation"},"content":{"rendered":"

HONOR has officially announced that its Magic V5<\/strong> will have the on-device multilingual call translation system<\/strong>\u2014a leap forward in both speed and accuracy. This innovation marks a significant milestone in multilingual speech recognition and translation, addressing long-standing technical challenges such as low-latency streaming recognition and the efficient deployment of large-scale models on mobile hardware<\/strong>.<\/p>\n

Privacy Meets Performance<\/h3>\n

Traditional translation tools often rely on the cloud, raising serious privacy concerns<\/strong>, particularly for sensitive conversations like phone calls. On-device solutions exist, but they often compromise speed, accuracy, or efficiency due to limitations in mobile hardware. HONOR\u2019s new technology changes the equation. By delivering cloud-level translation performance entirely on-device<\/strong>, it eliminates privacy risks while maintaining top-tier speed and accuracy.<\/p>\n

Smarter Storage and Real-Time Translation<\/h3>\n

One of the standout achievements is the memory efficiency<\/strong>. HONOR has slashed the footprint from 3\u20134GB to just 800MB<\/strong>, saving around 75% of storage space<\/strong>. Six language packs\u2014Chinese, English, German, French, Spanish, and Italian<\/strong>\u2014are embedded directly, avoiding the need for bulky downloads and saving roughly 2.78GB of storage<\/strong>.<\/p>\n

\"Honor<\/p>\n

The Magic V5 also introduces \u201cspeak-as-you-go\u201d real-time translation<\/strong>, moving beyond traditional methods that wait for full sentences. This improvement boosts inference speed by 38%<\/strong> and translation accuracy by 16%<\/strong>, making conversations more natural and seamless.<\/p>\n

Backed by INERSPEECH 2025 Research<\/h3>\n

HONOR\u2019s announcement is underpinned by peer-reviewed research presented at INTERSPEECH 2025<\/strong>.<\/p>\n

    \n
  • \n

    The first paper, MFLA: Monotonic Finite Look-ahead Attention for Streaming Speech Recognition<\/em>, details how HONOR adapted the Wait-k strategy<\/strong> (common in text translation) to speech recognition by combining it with a Continuous Integrate-and-Fire (CIF)-based predictor<\/strong>, drastically cutting latency while preserving accuracy.<\/p>\n<\/li>\n

  • \n

    The second paper, Novel Parasitic Dual-Scale Modelling for Efficient and Accurate Multilingual Speech Translation<\/em>, introduces a parasitic dual-scale speculative sampling approach<\/strong> (developed with Shanghai Jiao Tong University<\/strong>) that accelerates inference by 38%<\/strong> on-device without compromising quality.<\/p>\n<\/li>\n<\/ul>\n

    The Future of On-Device AI<\/h3>\n

    With the Magic V5, HONOR signals its commitment to private, real-time, AI-driven communication<\/strong>. By translating faster, lighter, and more secure\u2014without the cloud\u2014this breakthrough sets the stage for a new era of intelligent human-device interaction<\/strong>.<\/p>\n","protected":false},"excerpt":{"rendered":"

    HONOR has officially announced that its Magic V5 will have the on-device multilingual call translation system\u2014a leap forward in both speed and accuracy. This innovation marks a significant milestone in multilingual speech recognition and translation, addressing long-standing technical challenges such as low-latency streaming recognition and […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":false,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"jetpack_post_was_ever_published":false},"categories":[27,12,36],"tags":[8,298],"class_list":["post-232","post","type-post","status-publish","format-standard","hentry","category-ai","category-android","category-mobile","tag-honor","tag-magic-v5"],"jetpack_publicize_connections":[],"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=\/wp\/v2\/posts\/232","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=232"}],"version-history":[{"count":0,"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=\/wp\/v2\/posts\/232\/revisions"}],"wp:attachment":[{"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=232"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=232"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/technewscentury.co.uk\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=232"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}