HyperFlow open-sources data-free flow self-distillation on MiniMax H3, with a licensing catch
haider1 · x · 2026-10-20
Video Rebirth released HyperFlow, an open-source data-free flow self-distillation method built on MiniMax H3. The model generates its own candidate videos, filters them for quality, and self-distills from passing outputs — no external data or human labels. It's framed as the self-distillation component of an RSI (recursive self-improvement) training paradigm for video models.
Highlights: near-frontier quality at lower inference cost, smooth fast-motion handling, and support for NVIDIA's Sol-Attn sparse-attention kernel for up to 13% faster H200 generation than dense attention. Code is Apache-2.0 on GitHub with weights on Hugging Face — but the MiniMax H3 Community License excludes the US, EU, UK and South Korea, so check terms before downloading.
Related event: HyperFlow Open-Sources 3x Speedup for MiniMax H3 Video Generation(3 posts)→
More from Multimodal
- Customuse launches Smart Mesh AI with geometry-aware adaptive character clothing tool — SimplyAnnisa · 2026-10-23
- AI-generated movie poster 'The Chase' crams a wolf, a plane and muscle cars into one chase — Aggravating-Sir-115 · 2026-10-23
- SteerTTS-edge: an 11.84MB steerable TTS with unchanged WER running 19x realtime on CPU — Working_Hat5120 · 2026-10-23
- ElevenLabs' Chimera pairs Eleven v4 voice with Seedance video to make character images speak — AIwithGhotai · 2026-10-23
- Open-Source Tennis Models Plus Claude Turn Video Into a World-Class Coach — tweetsatpreet · 2026-10-23
- Berkeley's StyleStream 2.0 cuts streaming voice conversion latency to 520ms, 2x faster — Berkeley · 2026-10-23