Startup Underdog launched Woof, a 31.7MB text-to-speech model with 31 million parameters that runs entirely in a browser without internet access. The company showcased it through a retro-themed browser game where users can type text for the model to speak and manipulate a music tuner, all voiced by Woof. Underdog made the model open for download, and the demo follows the company's earlier release of Bark, a 167MB speech-to-text model for edge devices.
techcrunch.com
· 2026-11-06
Ollama has added text-to-speech support to its Mac app, starting with Phonon, a family of small speech models from Gradium. The feature generates audio in real time as text is produced, and Ollama says Windows and Linux support is coming soon. Ollama reports its MLX-based implementation runs 35% faster than Gradium's own version in tests on an M5 Max MacBook Pro.
9to5mac.com
· 2026-10-26
Destined AI, a small Atlanta-based team, has released Hurston-1, an open-source text-to-speech model fine-tuned from Maya1 to handle Southern American, Nigerian, West African, East African, and Southern African English accents. The model includes a narrowband phone mode producing 8 kHz G.711 audio compatible with Twilio and other telephony systems, alongside a default 24 kHz full-audio mode. It was trained partly on consented, paid recordings collected from contributors since 2021.
huggingface.co
· 2026-10-25
Block has introduced an AI voice model called Dessa, which demonstrates a range of human-like synthetic voices. The company shared audio samples showcasing the model's ability to mimic natural speech patterns and tones.
gizmodo.com
· 2026-10-24
Mistral AI has released Voxtral 3, a family of speech-to-text and text-to-speech models paired with a voice agent SDK that combines listening, reasoning, tool-calling and speaking in one real-time pipeline. The system supports over 50 languages and is designed for multi-step conversational tasks like contact-center support and hands-free assistants. It's available through Mistral's EU-hosted API or for self-deployment on customer infrastructure.
mistral.ai
· 2026-10-20
A Show HN project called PDF TO MP3 lets users convert PDF documents into audiobooks by connecting their Codex or Claude account via an MCP server. Document processing is covered by the user's existing AI subscription, while the actual text-to-speech generation is billed separately through an OpenAI or Google Gemini API key. Users authenticate the MCP server, add their speech API key, then attach a PDF in their chat client and request an audiobook conversion.
pdftomp3.com
· 2026-10-13
Suno has released Speech, a public beta feature on its web and mobile apps that generates spoken voiceovers alongside background music as a single track. Users can create audio via a Simple mode, describing a scene in a prompt, or an Advanced mode that accepts a custom script and lets them adjust voice gender, style and variation; background music can also be toggled off.
theverge.com
· 2026-10-02
Translation earbuds, such as Apple's AirPods Pro paired with the Translate app, let users hear real-time spoken translations by connecting to a phone, selecting languages and activating a live translation mode. Some designs require sharing an earbud with the other speaker instead. Behind this simple interface, the devices run three technical stages: speech recognition, machine translation, and text-to-speech synthesis.
engadget.com
· 2026-09-29
Google has launched Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, new speech generation models supporting over 100 languages. The models ranked first and second respectively on Hume AI's Overall Quality Index, with Gemini 3.8 Flash TTS also topping Hume AI's Voice Design Benchmark and its accent modeling category. Google says the models improve on long-form content and dual-speaker screenplay control compared to the prior Gemini 3.1 Flash TTS.
blog.google
· 2026-09-23
Nari Labs announced its Qwen3-TTS and Qwen3-ASR models achieved leading positions on Coval's voice AI benchmark, which measures latency and word error rate for speech AI systems. The ASR Fast model hit 44ms latency with 3.6% error rate at $0.12/hour, while the TTS Fast model achieved 63ms latency with a best-in-class 3.8% error rate, undercutting competitors like AssemblyAI and Deepgram on price.
narilabs.com
· 2026-09-14
Indian audio storytelling platform Pocket FM has doubled its annualized revenue run rate to $500 million over the past year, driven largely by AI-generated content. CEO Rohan Nayak said AI now accounts for 93% of the platform's overall catalog and 99% of new releases, though humans still shape the underlying story ideas.
techcrunch.com
· 2026-09-10
GoKawiil