Artificial Analysis released a TTS accent benchmark showing that accent materially changes model rankings. Qwen-Audio-3.1-TTS-Plus leads in US English, while HeyGen Voice performs better in UK English and tops the overall leaderboard.
Text-to-Speech (TTS)
语音合成
Technology that converts text into natural-sounding human speech.
Curated Intelligence
OpenRouter announced that HeyGen Voice Turbo is free until November 30. The text-to-speech model supports 39 languages and over 2,900 voice options, handling up to 5,000 characters per request with streamed 44.1 kHz audio. This lower-latency version previously topped the Artificial Analysis TTS leaderboard.
Voice AI firm Deepgram has acquired Daily, the developer of the popular open-source framework Pipecat, aiming to consolidate real-time voice and video infrastructure capabilities. While the deal amount remains undisclosed, both companies committed to keeping Pipecat open-source. This move highlights the accelerating consolidation within the voice AI infrastructure sector.
AssemblyAI has released its first text-to-speech model, AssemblyAI-TTS-1, now available on OpenRouter. Designed for voice agents, it offers streaming with 90ms latency and over 4,000 voices at $15 per million characters. Third-party benchmarks rank it highly in speed, accuracy, and naturalness.
AssemblyAI has released its first Text-to-Speech model, AssemblyAI-TTS-1, ranking #12 in Provider Voice and #10 in Controlled Voice benchmarks. The model demonstrates strong pronunciation robustness and offers a highly competitive pricing structure compared to top-tier rivals.
Voice Arena launched five new TTS leaderboards for languages including French and German, using blind voting by native speakers. ElevenLabs models ranked first across all five boards, leading the next-best provider by 11 to 26 Elo points.
Ant Group has released and open-sourced Ming-3.0-tts, a text-to-speech model with only 0.5B parameters under the MIT license. It ranks in the top 3 of Artificial Analysis's TTS leaderboards, becoming the only sub-1B parameter model to achieve this tier, offering 7x to 70x fewer parameters than other leading open-weight TTS models.
HeyGen released HeyGen Voice Turbo, which secured the #1 spot on the Artificial Analysis Controlled Voice Arena Leaderboard. The model offers significant cost efficiency at $4/1M characters and demonstrates improved pronunciation robustness and accent handling compared to its predecessor and competitors.
Mistral AI has launched the Voxtral 3 series, featuring two new speech-to-text models and one text-to-speech model supporting over 50 languages. Designed to power Mistral's new voice agents API, these models are now available on OpenRouter for developer access.
HeyGen's new text-to-speech model, HeyGen Voice, is now available on OpenRouter after topping the Artificial Analysis TTS leaderboard in blind tests. The service offers standard and low-latency Turbo modes, with API users receiving a 50% discount ($15 per million characters) through October 31.