Neuronaut Releases 35th Collection of Tools for MiniMax-H3

The 'Neuronaut' channel released its 35th collection of open-source tools for the MiniMax-H3 video model — ranging from character references via RefMod (Prompt Rewriter 0.32) to accelerators like Rapid-H3 and RelayHires; the open model base on Hugging Face has surpassed 6.1 million downloads.

Compiled by Sergey KostenchukPublished 2026-11-02Updated 2026-11-02

2026-11-02 Coding NVIDIA

🛠 Neuronaut Releases 35th Collection of Tools for MiniMax-H3

The main highlight this time is Prompt Rewriter 0.32 by @Aleks_101: support for RefMod — a ready-made character reference (appearance and voice) from a file. The open MiniMaxAI/MiniMax-H3 base on Hugging Face has surpassed 6.1 million downloads.

🌍 The base model remained unchanged, but the ecosystem addressed several stages of the production cycle: 3–5x acceleration on DGX Spark via INT8 cores (Rapid-H3, around 2 minutes per clip), anime video upscaling from 0.5 to 2 MP without turbo-LoRA (RelayHires), and serial assembly of episodes with 280–300 shots.

👤 Ready-made recipes: slow generation — Rapid-H3 presets; soapy faces — second pass of Face Refiner with denoise 0.4 (around 24 minutes on RTX 3060); character from file — RefMod (it slows down generation); long episodes — seriesmith or FrameForge.

Source 1: https://github.com/pytraveler/MiniMax-H3-Prompt-Rewriter-ComfyUI Source 2: https://github.com/jeet0733/Rapid-H3

Sources