News linked to both this project and an event.
Microsoft has removed the anthropomorphic chatbot Mico from Copilot voice features and moved it to the Learn Live learning center to serve as a "tutor". The support page states that the interaction experience accumulated by Mico is shaping the future direction of Copilot, and Copilot voice interaction features remain available. Mico was launched in October 2025, enabled by default in voice mode, and previously included an Easter egg where clicking turns it into Clippy.
According to Cryptopolitan, South Korea's Financial Supervisory Service (FSS) has launched a three-month fraud refund system reconstruction project with a budget of approximately 119 million Korean won. This upgrade aims to align with the revised "Telecommunications Fraud Damage Compensation Act" amended on March 31 this year—the bill will officially take effect on October 1, formally incorporating virtual assets into the scope of "damaged property" and "refundable property." The new system will support calculating compensation amounts based on token type and quantity, using the Korean won value at the time funds were frozen as a reference benchmark, while also possessing the ability to disentangle mixed fraud funds across multiple accounts. Major exchanges such as Upbit, Bithumb, Coinone, Korbit, and GOPAX will be required to assume anti-fraud and victim relief obligations equivalent to those of banks at that time, including verifying transaction purposes, monitoring suspicious funds, and freezing suspected accounts.
ByteDance Seed Team released SeedRealtime, a native audio-video full-duplex large model, now available on the Doubao App. The model unifies audio, video, and text within an end-to-end architecture, processing perception, understanding, decision-making, and expression in parallel, and internalizes turn-taking within the model, replacing external Voice Activity Detectors. Internal human evaluation at ByteDance shows that its rhythm issues are halved compared to cascaded architectures, but no technical report, parameter scale, open-source weights, or API have been released.
The Qwen App and PC client have launched new features, introducing multiple new functions such as Thinking & Research, Scheduled Tasks, Office Assistant, Agent Marketplace, and Voice Calls, while supporting Alibaba's latest flagship model Qwen3.8-MAX. All users can use the above features for free.
Tech blogger @synthwavedd has stated that several highly anticipated cutting-edge AI models have adjusted their release schedules. Among them, GPT-5.6, originally slated for a near-term release, has been postponed to mid-July, while Google DeepMind has canceled the planned late-June launch of Gemini 3.5 Pro due to dissatisfaction with the model’s current performance.Meanwhile, preparations for the launch of OpenAI’s new-generation bidirectional voice model, Bidi, are progressing on the ChatGPT platform, with availability to users potentially coming as early as this week. Reports indicate that Bidi supports full-duplex voice interaction, allowing users to speak simultaneously with the model and interrupt the conversation at any time, marking a significant upgrade over existing voice modes.Additionally, Anthropic has granted early access to its Claude Sonnet 5 model for select enterprise clients. As the release progress of flagship models Mythos 5 and Fable 5 has slowed, Claude Sonnet 5 is seen as an interim product solution for Anthropic to alleviate competitive pressure.
Odaily Odaily News: Voice AI company ElevenLabs, founded in Poland, has released Music v2. This model can switch genres within the same track, generate songs by sections such as intro, verse, and chorus, and supports local redrawing of specified segments. Music v2 is now available on ElevenMusic and ElevenCreative, with early access to the ElevenAPI currently provided through the sales team. Stability AI has launched Stable Audio 3.0, which includes four models: Small SFX, Small, Medium, and Large. Three of these variants have open weights on Hugging Face. The series is trained on licensed data. The Medium model supports generating tracks up to 6 minutes and 20 seconds long, while the Large model is only available via the Stability AI API. ElevenLabs has signed licensing agreements with Believe, Kobalt, and Merlin. Stability AI has established partnerships with Warner Music Group and Universal Music Group. In November 2025, Suno was valued at $2.45 billion, with an annual recurring revenue exceeding $300 million, approximately 100 million users, and generating an average of about 7 million songs daily.
According to the official OpenClaw GitHub release notes, OpenClaw released version v2026.4.25 on April 27. This update covers multiple core modules—including speech, plugins, observability, and browser automation—with the following key highlights: • TTS Overhaul: Added support for six new voice service providers—Azure Speech, Xiaomi MiMo, ElevenLabs v3, Inworld, Volcengine (Volcano Engine), and local CLI—and introduced chat-session-level automatic TTS control, TTS persona binding, and cross-Agent/account-level override configurations. • Plugin Registry Optimization: Migrated plugin startup and installation paths to a cold-persistent registry, reducing full-manifest scanning and improving determinism for plugin updates, fixes, and discovery. • OpenTelemetry Expansion: Extended coverage across the full observability pipeline—including model invocations, token usage, tool loops, Harness execution, process execution, message delivery, context assembly, and memory pressure metrics. • Browser Automation Enhancements: Added iframe-aware role snapshots, CDP readiness tuning, one-click headless startup, and deeper browser diagnostic probes. • Installation Stability Improvements: Fixed installation and update issues across Windows, macOS, Linux, and Docker platforms.
Odaily News According to Shanghai Internet Information Office, Shanghai has added one more registered generative artificial intelligence service, which is Tesla's in-car voice large model service. In response, a Tesla representative stated on April 22nd: "Stay tuned." (Yicai)
According to threat intelligence released by the SlowMist security team (@SlowMist_Team), its threat intelligence system MistEye has received community reports identifying an active social engineering attack targeting cryptocurrency users. Attackers contact victims under the pretext of project collaboration and lure them into using a counterfeit “Harmony Voice” application (domain: harmony-voice[.]app) for so-called real-time translation—when in fact it is malicious software. SlowMist has already synchronized the relevant threat intelligence (IOCs) to its enterprise customers.