Techub News News, Zyphra releases open-source text-to-speech model ZONOS3-Mini, designed for real-time use on local devices.ZONOS3-Mini is a sparse mixture-of-experts model with a total parameter count of 8 hundred million, with per-token activated parameter count of 1 hundred million, supporting zero-shot voice cloning and word-level speech control. Its weights are open-sourced under the Apache 2.0 license. The model can run in a browser tab, on a laptop, or on a server GPU , supporting NVIDIA GPU, Apple Silicon,CPU/CUDA/Metal/Vulkan and via WebGPU Runs in a phone or browser. Zyphra says it ranks first among open-source models in its own ZTTS2-Eval capability index. ZONOS3-Mini uses a new audio codec ZAC-1 and a hybrid MoE backbone network, and integrates a new speaker encoder. Zyphra used 112 blocks NVIDIA H100 GPU trained for 2.5 days. (MarkTechPost)
All information provided on this website is for reference only.
This website does not guarantee the accuracy, validity, timeliness, or completeness of the information.
Any actions taken based on the information provided by this website are at the user's own risk.
如您认为本页面(或本平台)内容存在侵权或违规情形,请您在发现之日起尽快与我们联系,并提供:权利人姓名/名称及联系方式、被侵权作品名称及权属证明、具体侵权链接及情况说明。我们在核实后将及时处理相关内容。
联系方式:creator@techub.news
商务合作:contacting@techub.news
点击链接下载 Techub News App,获取最新讯息:
https://www.techub.news/download
(或扫描下方二维码下载)

Techub News News, Zyphra releases open-source text-to-speech model ZONOS3-Mini, designed for real-time use on local devices.ZONOS3-Mini is a sparse mixture-of-experts model with a total parameter count of 8 hundred million, with per-token activated parameter count of 1 hundred million, supporting zero-shot voice cloning and word-level speech control. Its weights are open-sourced under the Apache 2.0 license. The model can run in a browser tab, on a laptop, or on a server GPU , supporting NVIDIA GPU, Apple Silicon,CPU/CUDA/Metal/Vulkan and via WebGPU Runs in a phone or browser. Zyphra says it ranks first among open-source models in its own ZTTS2-Eval capability index. ZONOS3-Mini uses a new audio codec ZAC-1 and a hybrid MoE backbone network, and integrates a new speaker encoder. Zyphra used 112 blocks NVIDIA H100 GPU trained for 2.5 days. (MarkTechPost)
All information provided on this website is for reference only. The website does not guarantee the accuracy, validity, timeliness, or completeness of the information. Any actions based on the information provided are at the user’s own risk.
如您认为本页面(或本平台)内容存在侵权或违规情形,请您在发现之日起尽快与我们联系,并提供:权利人姓名/名称及联系方式、被侵权作品名称及权属证明、具体侵权链接及情况说明。我们在核实后将及时处理相关内容。
联系方式:creator@techub.news
商务合作:contacting@techub.news
点击链接下载 Techub News App,获取最新讯息:
https://www.techub.news/download
(或扫描下方二维码下载)
