{"name":"Haru","repo":"higurashi-ai/Haru","url":"https://midorreal.com/project/higurashi-ai-haru","repository":"https://github.com/higurashi-ai/Haru","description":"Haru: a low-latency, multilingual, full-duplex spoken dialogue model","facts":{"license":{"value":"Apache-2.0","source":"https://github.com/higurashi-ai/Haru","checked":"22 Nov 2026"},"language":{"value":"Python","source":"https://github.com/higurashi-ai/Haru","checked":"22 Nov 2026"},"homepage":{"value":"https://huggingface.co/higurashi-ai/Haru-4B","source":"https://huggingface.co/higurashi-ai/Haru-4B","checked":"22 Nov 2026"},"releases":{"value":"0","source":"https://github.com/higurashi-ai/Haru/releases","checked":"19 Nov 2026"},"contributors":{"value":"2","source":"https://github.com/higurashi-ai/Haru/graphs/contributors","checked":"19 Nov 2026"},"archived":{"value":"no","source":"https://github.com/higurashi-ai/Haru","checked":"22 Nov 2026"}},"stars":724.0,"writeup":{"what_it_is":"Haru is a full-duplex spoken dialogue model that listens and speaks at the same time. This repository contains the inference code for Haru-4B, a pretrained base model, with PyTorch, MLX and vLLM-Omni implementations and a web demo.","audience":"Researchers and developers working with real-time speech dialogue models.","claims":[{"kind":"specific","claim":"Processes one 133 ms frame at a time and produces one frame of audio and text per step.","excerpt":"Every 133 ms the model takes one frame of the user's audio and produces one frame of its own audio and text.","status":"not_checked"},{"kind":"specific","claim":"Supports nine languages.","excerpt":"English, Spanish, Portuguese, German, French, Chinese, Japanese, Russian and Korean.","status":"not_checked"},{"kind":"specific","claim":"Audio is a 64-dimensional latent per frame, generated in one sampling step.","excerpt":"Audio is a 64-dimensional latent per frame. A small MLP head with MeanFlow sampling produces it in one step","status":"not_checked"},{"kind":"specific","claim":"Runs on an NVIDIA GPU with about 12 GB of memory.","excerpt":"With an NVIDIA GPU (about 12 GB of memory):","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-14"},"why_now":null,"owner_supplied":null,"score":{"verdict":"too_early","hype":73,"reality":30,"gap":44,"momentum":null,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T07:00:00+00:00","verdict_source":"rule"},"early_read":{"estimate":true,"lean":"too_close","gap_low":9,"gap_high":50,"rule":"early-2.0"},"changes":[{"date":"2026-11-12T11:15:42+00:00","text":"Repository created","source":"https://github.com/higurashi-ai/Haru"}],"cite":"Haru is a GitHub project at https://github.com/higurashi-ai/Haru written mainly in Python, described there as \"Haru: a low-latency, multilingual, full-duplex spoken dialogue model\".","attribution":"Data from Mid or Real, https://midorreal.com/project/higurashi-ai-haru, checked 22 November 2026. Attribution with a link is required.","still_real":null}