Audio & Speech

Today's Stories

1. 1

   Google、Gemini 3.7 Flash公開、Pixel 11発表

   Google released Gemini 3.7 Flash, its latest developer model, and unveiled the new Pixel 11 series. The Gemini app officially surpassed 1 billion monthly users. Gemini 3.7 Flash arrived just three weeks after 3.6 Flash at half the introductory price per million tokens. The Pixel 11 devices run the Google Tensor G6 chip with the latest Gemini Nano model.

   Gemini Omni 1.1 Flash now supports scene extension, 4K upscaling, and faster prototyping. Gemini 3.5 Transcribe offers real-time transcription for voice agents and live captioning.

2. 2

   Phonely launches Alma, voice AI trained on 10M calls

   Phonely Ltd. launched Alma, a large language AI model built for voice agents and trained on over 10 million real phone conversations. Alma provides sub-185-millisecond response times to the first token, compared with roughly 500 milliseconds for OpenAI's GPT-4.1. Many AI models for voice agents are trained on text, which can make conversations sound unnatural. Alma reduces pauses that interrupt conversations and is pre-trained to sound like a person, cutting the need for prompt engineering. The model also costs 55 cents per blended million tokens, making it 84% cheaper than GPT-4.1 and 90% cheaper than GPT-5.4.

   Alma uses real traffic to improve how it interacts with calls without waiting for the next training cycle, aiming to deliver better results the next day. It can work with existing transcribers and text-to-speech providers, and is already answering millions of Phonely calls monthly.

3. 3

   TontaubeV1 released: character-level TTS for long-form

   Two developers released TontaubeV1, a 2.9B-parameter open-weight TTS model for expressive speech and long-form narration. It supports zero-shot voice cloning from up to one minute of reference audio and targets English and German. The model uses character-level tokenization, which the authors found works better than the original BPE tokenizer in early experiments. This choice, plus building on the DualCodec multi-codebook audio codec, aims to improve long-form generation and low-latency local inference.

   The model was trained on 7 languages and about 200k hours of audio, but is mostly tested in English and German. Availability of the open weights and the specific licensing or access details were not mentioned in the announcement.

4. 4

   Mitsubishi Electric develops task-general sound separation AI

   Mitsubishi Electric and its U.S. subsidiary Mitsubishi Electric Research Laboratories have developed a single AI model that can separate and extract specified sounds from mixed audio based on a text prompt, handling tasks like separating multiple speakers, speech enhancement, and extracting environmental sounds. This addresses a key limitation of earlier technology, which required a separate model for each separation task and struggled to adapt to different noise environments and use cases. The new capability is expected to improve AI systems' understanding of complex situations in factories and public spaces, boosting the reliability of physical AI (AI that interacts with the physical world).

   The technology will be demonstrated at CEATEC 2026 (October 13-16, at Makuhari Messe) as "sound source separation technology for distinguishing abnormal sounds and human voices from varied site sounds," with a live demo separating factory noises, musical instruments, and multiple voices.

5. 5

   Musician Detectives Hunt AI Music Grifters

   EDM producer Max "H4RRIS" Harris and Italian turntablist-turned-producer Nihil Young are publicly calling out AI-generated music in the EDM scene, often using tracks from platforms like Suno as examples. Harris and Young argue AI music is "decoy art" that steals from real artists, and Young says he's lost clients since AI rolled out, highlighting the financial harm to human creators.

   Young has taken a break from callouts after receiving hacking attempts and harassment, but AI music is growing—Deezer reports AI songs now account for over 50% of new uploads.

6. 6

   Beatport bans AI-generated music from DJ marketplace

   Beatport now bans tracks made entirely or mostly by AI. The rule is enforced with a detection tool from partner Beatdapp that filters AI tracks during upload, and rights holders receive a notice when a track is rejected. A Beatport survey found 60 percent of users would not play AI music in their sets, and 77 percent prefer human-made music. CEO Matt Gralen stressed the difference between a tool that helps people and a system that replaces them entirely.

   Other platforms like Deezer have recently taken similar steps, with Deezer noting that nearly 50 percent of its daily uploads are AI-generated.