Testing DGX Spark with H3: First Text-to-Video Run via ComfyUI
allaboutai-kris · reddit · 2026-08-11
The author shares their first experience running the H3 model on the DGX Spark. This test used a text-to-video (t2v) approach without reference images, controlling ComfyUI via Codex driven by GPT-5.6. Although the final video output was far from perfect, the author believes this workflow represents the future of AI video generation.
More from Infra
- Meta Returns to Open Source: Muse Glimmer Hits 230 tok/s on RTX 5090 via SGLang — ying11231 · 2026-08-11
- ZEISS Creates the Smoothest Mirrors Ever: Hardcore Specs of EUV Lithography — DynamicWebPaige · 2026-08-11
- Running Open-Source LLMs on Dual AMD v620: Tensor Split Nearly Doubles Speed — Thin_Pollution8843 · 2026-08-11
- NVIDIA Partners with Wall Street Giants to Mobilize $500 Billion for AI Infrastructure — borowcy · 2026-08-11
- UBS Bullish on AI PCB Cycle: Q-glass on Hold, M9/PTFE Hybrid Emerges — zephyr_z9 · 2026-08-11
- Running Meta's Muse Glimmer 30B on MacBook: Ollama MLX Hits 29 tok/s — ollama · 2026-08-11