←Search results

Search “Speech”

By relevance · Total 1394 results Clear search

Friday

  1. Story2 reports · 2 facts

    Neuralink team develops 8 brain-computer interface features in 72 hours

    In a 72-hour hackathon, the Neuralink team developed 8 functional prototypes for P21 participant Kenneth, including silently thinking "Hey Neuralink" in his mind and thinking through a task to trigger the Grok agent to operate a computer autonomously. The team also rewrote the neural voice activity detection system, achieving pure thought-based speech decoding without the need for button presses, and routed the decoded speech into the iPhone call microphone audio stream, allowing this ALS patient to call his wife using only his thoughts. In addition, the dedicated app Bloom supports placing orders by thought in a café, and a system-level brain-computer input method allows text entry in any app. On the same day, Elon Musk announced that Neuralink's first product is named Telepathy, stating that its output bandwidth will far exceed the fastest human speech or typing speed within a few years; the tweet cited a post about Neuralink engineers conducting a 72-hour hackathon with a speech-restoration research participant and rapidly prototyping multiple thought-communication features.

    In a 72-hour hackathon, the Neuralink team developed 8 functional prototypes for P21 participant Kenneth, including silently thinking "Hey Neuralink" in his mind and thinking through a task to trigger the Grok agent to operate a computer autonomously. The team also rewrote the neural voice activity detection system, achieving pure thought-based speech decoding without the need for button presses, and routed the decoded speech into the iPhone call microphone audio stream, allowing this ALS patien

    Matched fieldsSummary

  2. Story4 reports · 1 facts

    ChatGPT launches SMS interaction feature for dot

    ChatGPT announces the launch of an SMS interaction feature for dot: users can text dot via iMessage or RCS, assign tasks using text, photos, or voice messages, and also have it keep an eye on something continuously, with dot proactively sending messages to report progress. The feature is being rolled out in preview to Pro users in the US. According to the official statement, only users themselves can message their own dot, and dot can only reply to the user; moreover, dot and OpenAI will not be given access to users' other SMS conversations. To set it up, select Texting on dot's profile page; dot shares the same context and plugins across different entry points.

    ChatGPT announces the launch of an SMS interaction feature for dot: users can text dot via iMessage or RCS, assign tasks using text, photos, or voice messages, and also have it keep an eye on something continuously, with dot proactively sending messages to report progress. The feature is being rolled out in preview to Pro users in the US. According to the official statement, only users themselves can message their own dot, and dot can only reply to the user; moreover, dot and OpenAI will not be

    Matched fieldsSummary

  3. Story1 reports · 1 facts

    Google Releases EmbeddingGemma 2 Paper

    Google has released the EmbeddingGemma 2 paper, introducing an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. Built on Gemma 4, the model learns from the larger Gemini Embedding 2 and is tuned with contrastive training; with 744M parameters, it outperforms models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal model takes about 567MB; the context window is expanded 4 times to 8K tokens, accommodating about 5.5 minutes of audio, 29 images, or 58 video frames. Embedding speech directly scores 69.54 on the MSEB spoken query retrieval benchmark, higher than the 57.59 achieved by transcribing first and then retrieving, and code retrieval on MTEB Code improves from 68.76 in EmbeddingGemma 1 to 78.68. There are currently no further developments.

    Google has released the EmbeddingGemma 2 paper, introducing an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. Built on Gemma 4, the model learns from the larger Gemini Embedding 2 and is tuned with contrastive training; with 744M parameters, it outperforms models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal

    Matched fieldsSummary

  4. Story5 reports · 1 facts

    Anthropic upgrades Claude mobile iPad and iPhone Duo interface

    On October 30, 2026, Anthropic upgraded the Claude mobile app; on iPad and iPhone Duo, generated content now opens alongside the chat. In the same batch of updates, the iOS and Android versions support simultaneous login to multiple accounts with switching without logging out; Chromebook, Googlebook, and foldable devices gain an adjustable sidebar, touchpad and right-click support, and keyboard shortcuts; voice mode on iOS and Android responds faster and recognizes speech more accurately in noisy environments, and iPhone voice mode is integrated with Apple CarPlay. Phoenix Technology's report also mentioned that Claude added 24 new languages, bringing cumulative support to 33 languages.

    On October 30, 2026, Anthropic upgraded the Claude mobile app; on iPad and iPhone Duo, generated content now opens alongside the chat. In the same batch of updates, the iOS and Android versions support simultaneous login to multiple accounts with switching without logging out; Chromebook, Googlebook, and foldable devices gain an adjustable sidebar, touchpad and right-click support, and keyboard shortcuts; voice mode on iOS and Android responds faster and recognizes speech more accurately in nois

    Matched fieldsSummary

  5. Story1 reports · 1 facts

    Joy releases a 100,000-hour unscripted two-person dialogue dataset

    On October 30, 2026, Shobhit Banga released Joy, a dual-track dialogue dataset containing 100,000 hours of real, unscripted two-person conversations, with each person on a separate audio track, covering 50,000 speakers and 10 languages. In a report the same day, Rohan Paul noted that current speech AI is mainly trained on fake conversations written by LLMs and read aloud by TTS, in which people cannot be heard interrupting each other, and he was curious about how these conversations are sustained for 10 to 20 minutes. The dataset has now been released, with no subsequent development progress yet.

    On October 30, 2026, Shobhit Banga released Joy, a dual-track dialogue dataset containing 100,000 hours of real, unscripted two-person conversations, with each person on a separate audio track, covering 50,000 speakers and 10 languages. In a report the same day, Rohan Paul noted that current speech AI is mainly trained on fake conversations written by LLMs and read aloud by TTS, in which people cannot be heard interrupting each other, and he was curious about how these conversations are sustaine

    Matched fieldsSummary

  6. Story1 reports · 1 facts

    VEED Dubbing launches on the fal platform

    On October 28, 2026, fal announced that VEED Dubbing is now available on the fal platform. According to information released by fal, this feature can translate videos using cloned voices and lip sync; the Pro version supports 94 languages and 14 regional accents, preserves background audio and supports multiple speakers, and also provides controls for target accent and voice clone strength. There are currently no subsequent development facts.

    Matched fieldsSummary

  7. Story1 reports · 1 facts

    OpenBMB showcases community open-source ESP32-S3 simulated tower dialogue project

    On October 29, 2026, OpenBMB showcased an ESP32-S3 cockpit-style attitude indicator project open-sourced by OpenBMB ambassador @emwstudio. The project uses MiniCPM5-2B to generate tower replies and VoxCPM2 to synthesize speech; the device sends push-to-talk audio over Wi-Fi to a Mac-side pipeline, where Qwen3-ASR performs local recognition, and the reply audio then streams back to the device speaker. The tower combines simulated flight data such as altitude, speed, heading, pitch and roll to point out inconsistencies between requests and instrument readings. The project is currently in an open-source showcase state, with no further progress disclosed.

    On October 29, 2026, OpenBMB showcased an ESP32-S3 cockpit-style attitude indicator project open-sourced by OpenBMB ambassador @emwstudio. The project uses MiniCPM5-2B to generate tower replies and VoxCPM2 to synthesize speech; the device sends push-to-talk audio over Wi-Fi to a Mac-side pipeline, where Qwen3-ASR performs local recognition, and the reply audio then streams back to the device speaker. The tower combines simulated flight data such as altitude, speed, heading, pitch and roll to poi

    Matched fieldsSummary

  8. Story1 reports · 1 facts

    Sarvam AI promotes Indus Cowork to 2,500 officials in Maharashtra

    Sarvam AI is rolling out its Indus Cowork tool to 2,500 government officials in Maharashtra for drafting documents, analyzing reports, summarizing files and retrieving government records. According to Analytics India Magazine, the tool was launched on October 28 by Chief Minister Devendra Fadnavis. The state approved Sarvam's sovereign AI proposal in August this year, with a two-year budget of 1.126 billion rupees; besides Indus Cowork, it also includes Samvaad, a voice and WhatsApp system for citizen communication, and integrates the Saaras speech recognition and Bulbul text-to-speech models into the state's MahaAI platform. The platform supports 22 languages including Marathi, and offers tools such as GR Expert for retrieving government resolutions, email summarization and RFP bid evaluation; Sarvam says the system runs on securely audited infrastructure located within India.

    Sarvam AI is rolling out its Indus Cowork tool to 2,500 government officials in Maharashtra for drafting documents, analyzing reports, summarizing files and retrieving government records. According to Analytics India Magazine, the tool was launched on October 28 by Chief Minister Devendra Fadnavis. The state approved Sarvam's sovereign AI proposal in August this year, with a two-year budget of 1.126 billion rupees; besides Indus Cowork, it also includes Samvaad, a voice and WhatsApp system for c

    Matched fieldsSummary

  10. Story1 reports · 1 facts

    Google releases second October update for Gemini for Home

    Google releases second October update for Gemini for Home, fixing an issue where Home and Nest speakers and displays would randomly say "Sorry, I didn't understand" when no one had asked a question. This update also fixes the reliability of YouTube TV voice commands on smart displays, an issue where adjusting the volume across rooms would mistakenly change the current device, and an issue where alarm and timer volumes were set to zero. The update is rolling out starting today.

    Matched fieldsSummary

  11. Story1 reports · 1 facts

    Apple Watch Series 12 and Ultra 4 to Add Siri Recap and Live Rewind

    9to5Mac reports that Apple Watch Series 12 and Apple Watch Ultra 4 will gain two new features, Siri Recap and Live Rewind, later this year via watchOS 27.3. Siri Recap generates conversation summaries using Apple Intelligence, and unsaved summaries are automatically deleted after 7 days; Live Rewind displays a text transcript of the past 15 seconds of speech when the Digital Crown is double-clicked. There is currently no further development or official response.

    Matched fieldsSummary

  12. Story1 reports · 1 facts

    Deepgram announces acquisition of Daily, the developer of Pipecat

    Deepgram announced the acquisition of Daily, the developer of the open-source voice AI framework Pipecat, with the transaction amount undisclosed; both parties pledged that Pipecat will remain open source. According to TechCrunch AI, Daily had previously raised more than $62 million in cumulative funding; earlier this year, Deepgram became a unicorn with a $130 million Series C round. Deepgram also released the Real-Time Context Standard (RCS), an open protocol developed jointly with the Pipecat team, used to define metadata such as audio context, model instructions, and reply tone in voice agents. No further developments have been disclosed at this time.

    Deepgram announced the acquisition of Daily, the developer of the open-source voice AI framework Pipecat, with the transaction amount undisclosed; both parties pledged that Pipecat will remain open source. According to TechCrunch AI, Daily had previously raised more than $62 million in cumulative funding; earlier this year, Deepgram became a unicorn with a $130 million Series C round. Deepgram also released the Real-Time Context Standard (RCS), an open protocol developed jointly with the Pipecat

    Matched fieldsSummary

  13. Story1 reports · 1 facts

    ElevenLabs Signs MoU with the Government of Karnataka

    At its first India summit, ElevenLabs signed a memorandum of understanding with the Government of Karnataka to explore the use of voice AI for citizen services, voice security, and voice restoration. The collaboration will begin with pilots in grievance handling, public hotlines, and welfare applications, covering Kannada and the 13 other Indian official languages already supported by the platform. ElevenLabs will also deploy a safety framework covering deepfake detection, watermarking, authorization, and restrictions on voice cloning, and will provide eligible residents with free lifetime voice reconstruction licenses through the One Million Voices program.

    At its first India summit, ElevenLabs signed a memorandum of understanding with the Government of Karnataka to explore the use of voice AI for citizen services, voice security, and voice restoration. The collaboration will begin with pilots in grievance handling, public hotlines, and welfare applications, covering Kannada and the 13 other Indian official languages already supported by the platform. ElevenLabs will also deploy a safety framework covering deepfake detection, watermarking, authoriz

    Matched fieldsSummary

  14. Story1 reports · 1 facts

    ElevenLabs releases the forced alignment model Cue

    On October 28, 2026, ElevenLabs released the forced alignment model Cue, which can pinpoint the exact time point for every character in audio and has already been used in the Scribe speech-to-text model family. According to the ElevenLabs blog, Cue covers more than 90 benchmark languages, can process recordings of any length in a single pass, and takes about half a second to align one hour of audio on production GPUs; on benchmarks across 29 languages, its accuracy is higher than that of all alignment tools tested by ElevenLabs. The model is also available separately through the Forced Alignment API. There is currently no information on further developments.

    On October 28, 2026, ElevenLabs released the forced alignment model Cue, which can pinpoint the exact time point for every character in audio and has already been used in the Scribe speech-to-text model family. According to the ElevenLabs blog, Cue covers more than 90 benchmark languages, can process recordings of any length in a single pass, and takes about half a second to align one hour of audio on production GPUs; on benchmarks across 29 languages, its accuracy is higher than that of all ali

    Matched fieldsSummary

Tuesday

  2. Story5 reports · 1 facts

    Apple releases iOS 27.2 and iPadOS 27.2

    Apple releases iOS 27.2 and iPadOS 27.2. According to MacRumors Apple Intelligence, in iOS 27.2, images generated by Image Playground and some photos edited via Apple Intelligence will embed Google's SynthID watermark, which marks AI-generated content at the pixel level without affecting image quality and is designed to withstand cropping, filters and compression; in testing, images edited with Extend and Reframe carried the watermark, while small-scale Clean Up edits did not add it and only larger edits did, and there is no detection entry point on iPhone, so it must be checked via the SynthID Detector website. Separately, according to heise online, iOS, iPadOS and macOS 27.2 expand language support for Siri AI, adding French, Spanish, Portuguese, Japanese and Korean, with Norwegian, Turkish, Traditional Chinese and Vietnamese to be supported before the end of the year, while German will not arrive until 2027 at the earliest; Siri AI is still labeled as Beta, and version 27.2 will support remembering previous conversations, deeper use of personal data, and inclusion of photos, calendar invitations and voicemails in iMessage, and Siri AI remains blocked on iPhone, iPad and Apple Watch in EU member states.

    Apple releases iOS 27.2 and iPadOS 27.2. According to MacRumors Apple Intelligence, in iOS 27.2, images generated by Image Playground and some photos edited via Apple Intelligence will embed Google's SynthID watermark, which marks AI-generated content at the pixel level without affecting image quality and is designed to withstand cropping, filters and compression; in testing, images edited with Extend and Reframe carried the watermark, while small-scale Clean Up edits did not add it and only lar

    Matched fieldsSummary

  3. Story1 reports · 1 facts

    Vapi launches traffic splitting feature

    On October 26, 2026, Vapi launched a traffic splitting feature that can allocate inbound and outbound call traffic to different versions of voice agents in proportions such as 0.1%, 1%, or 5%, before full release after validation. The feature supports version stickiness based on caller identity, and the official recommendation is to first do a two-way split, changing only one variable at a time. Previously, releasing a new version would immediately switch 100% of traffic over, so teams were afraid to iterate frequently. The feature is currently available, with no further development information yet.

    On October 26, 2026, Vapi launched a traffic splitting feature that can allocate inbound and outbound call traffic to different versions of voice agents in proportions such as 0.1%, 1%, or 5%, before full release after validation. The feature supports version stickiness based on caller identity, and the official recommendation is to first do a two-way split, changing only one variable at a time. Previously, releasing a new version would immediately switch 100% of traffic over, so teams were afra

    Matched fieldsSummary

Wednesday

  1. Story1 reports · 1 facts

    Paris Smaragdis proposes to make machines hear better than humans

    MIT professor Paris Smaragdis's research group proposes to give machines auditory capabilities surpassing those of humans, and to apply this goal to scenarios such as speech recognition, music understanding, and environmental monitoring. His research covers machine learning for sound source separation, intelligent devices that can augment human auditory environments, and tools that help people understand, manipulate, analyze, and create music. Smaragdis joined MIT in 2025 and holds appointments in the Music and Theater Arts program and the Department of Electrical Engineering and Computer Science. This direction is still at the research advancement stage, and no subsequent progress has been disclosed.

    MIT professor Paris Smaragdis's research group proposes to give machines auditory capabilities surpassing those of humans, and to apply this goal to scenarios such as speech recognition, music understanding, and environmental monitoring. His research covers machine learning for sound source separation, intelligent devices that can augment human auditory environments, and tools that help people understand, manipulate, analyze, and create music. Smaragdis joined MIT in 2025 and holds appointments

    Matched fieldsSummary

  2. Story1 reports · 1 facts

    Brilliant Labs launches $399 Halo AI glasses

    Singapore-based manufacturer Brilliant Labs released the Halo AI glasses on local time October 27, priced at $399. The device features a full-color Micro-OLED display, is equipped with the self-developed AI assistant Noa, supports real-time voice interaction and long-term memory, and is built around three experiences: Ask, Make, and Play, among which the Narrative system can turn real-world activities into interactive games that can be adjusted in real time. Both hardware and software adopt an open-source design, with on-device AI processing at the core; the device weighs slightly more than 40 grams, and the company claims a maximum battery life of 14 hours. Separately, according to a Unite.AI report, Noa and miniapps have daily usage limits, with Noa Plus at $9.99 per month and Noa Pro at $39.99 per month; the company also announced an investment from Goodwater Capital, without disclosing the amount.

    Singapore-based manufacturer Brilliant Labs released the Halo AI glasses on local time October 27, priced at $399. The device features a full-color Micro-OLED display, is equipped with the self-developed AI assistant Noa, supports real-time voice interaction and long-term memory, and is built around three experiences: Ask, Make, and Play, among which the Narrative system can turn real-world activities into interactive games that can be adjusted in real time. Both hardware and software adopt an o

    Matched fieldsSummary

  3. Story1 reports · 1 facts

    Astropad releases Workbench 1.4 update

    Astropad released the Workbench 1.4 update on October 27, 2026. According to 9to5Mac, this version adds a built-in terminal, allowing users to switch between the remote desktop and the terminal without reconnecting; terminal sessions persist across devices and support voice input of commands. It also adds native support for iPhone Duo and supports drag-and-drop of files between Macs within terminal sessions. Workbench is free for 20 minutes per day, with unlimited access costing $14.99 per month or $79.99 per year, and requires macOS 15, iOS 26, or iPadOS 26 or later. There is currently no information on further developments.

    Astropad released the Workbench 1.4 update on October 27, 2026. According to 9to5Mac, this version adds a built-in terminal, allowing users to switch between the remote desktop and the terminal without reconnecting; terminal sessions persist across devices and support voice input of commands. It also adds native support for iPhone Duo and supports drag-and-drop of files between Macs within terminal sessions. Workbench is free for 20 minutes per day, with unlimited access costing $14.99 per month

    Matched fieldsSummary