{"rewrite":{"id":"r_9c5ec1c561a4b45fc4477332","clusterId":"c_cf9252242ab2181a904e768e","slug":"shizuku-ai-runs-an-on-device-iphone-voice-demo-without-the-cloud","model":"deepseek-v4-1-flash","headline":"Shizuku AI Runs an On-Device iPhone Voice Demo Without the Cloud","summary":"Shizuku AI Technology Department released a prototype demo on November 5 in which the AI VTuber Shizuku holds voice conversations using only an iPhone. Speech recognition, replies, and voice synthesis all run on the device. The demo uses a lightweight on-phone language model, and the department says its conversational ability is lower than the usual Shizuku. On November 8 the department announced Shizuku-Duplex, a Japanese voice dialogue model that processes audio every 80 milliseconds.","whyItMatters":"The point of the experiment is the tradeoff it makes visible: dropping the cloud gets Shizuku onto a phone, but the model small enough to run there talks worse than the streamed version.","webCardHtml":"\u003cp\u003eThe demo swaps typed comments for speech. A user talks to the phone, recognition turns the audio into text, a language model on the device drafts a reply, and voice synthesis reads it back in Shizuku\u0026#39;s voice. The iPhone\u0026#39;s GPU and Neural Engine, the chip\u0026#39;s AI accelerator, are used to shorten response times.\u003c/p\u003e\u003cp\u003eShizuku AI Technology Department framed the prototype as a question about scale. As LLMs grow larger, staff asked how naturally a conversation could be held on a smartphone alone, and accepted a smaller brain for the test. The department described the result as a clumsier Shizuku than usual.\u003c/p\u003e\u003cp\u003eThe Duplex model announced on November 8 targets a different problem. Full-duplex processing lets it listen while it speaks, and complex searches are pushed to a separate model behind the scenes.\u003c/p\u003e","blueskyPost":"Shizuku AI's iPhone demo trades conversational quality for offline privacy. The department admits the on-device model talks worse than Shizuku, a gap it is now narrowing with Shizuku-Duplex's 80ms audio processing.","twitterPost":"Shizuku AI's iPhone demo runs offline but the department admits its talk is weaker than Shizuku. Shizuku-Duplex's 80ms processing is the fix.","threadsPost":"Shizuku AI's iPhone demo keeps everything local, but the department says the on-device model converses worse than the usual Shizuku. Shizuku-Duplex, announced November 8, processes audio every 80 milliseconds and targets that quality gap.","newsletterBlurb":"Shizuku AI Technology Department put the AI VTuber Shizuku on an iPhone with no cloud in the loop: speech recognition, language model replies, and voice synthesis all run on the device. The department says the on-phone model is smaller and less capable than the usual Shizuku. A separate Japanese voice dialogue model, Shizuku-Duplex, was announced on November 8 and processes audio every 80 milliseconds.","attributionJson":"[{\"source\":\"ASCII.jp\",\"url\":\"https://ascii.jp/elem/000/004/449/4449767/?rss\",\"title\":\"AI VTuber \\\"Shizuku\\\" Inside an iPhone? Experiment Letting Her Chat Without the Internet Released\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5570,"outputTokens":794,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1794208713,"createdAt":"2026-11-09T07:14:30.000Z","publishedAt":"2026-11-09T07:17:29.000Z","updatedAt":"2026-11-09T07:17:29.000Z"},"cluster":{"id":"c_cf9252242ab2181a904e768e","canonicalTitle":"AI VTuber「しずく」がiPhoneの中に？　ネットに頼らずおしゃべりできる実験が公開","representativeArticleId":"a_37f088c701fe131e77d04d0f","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"vtuber\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-11-09T06:30:00.000Z","lastSeenAt":"2026-11-09T06:30:00.000Z","updatedAt":"2026-11-09T07:17:26.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/449/4449767/?rss","title":"AI VTuber「しずく」がiPhoneの中に？　ネットに頼らずおしゃべりできる実験が公開"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"news","domain":"vtuber","is_roundup":false},"keyFacts":["Shizuku AI Technology Department released the on-device voice demo on November 5.","The prototype runs speech recognition, language model replies, and voice synthesis on an iPhone without the cloud.","The demo uses the iPhone's GPU and Neural Engine to speed up responses.","The department says the on-phone language model has lower conversational ability than the usual Shizuku.","Shizuku-Duplex, a Japanese voice dialogue model, was announced on November 8 and processes audio every 80 milliseconds."],"pageLinks":{"healthSlugs":[],"storySlugs":[],"entities":[],"stories":[],"entitiesOk":true,"storiesOk":true}}