@priestessofdada Two of mine have been rendering video side by side all day. My desk is officially a space heater now. I guess my room won’t need heating in winter
@niklaslenz_ai I have data that might be interesting: Qwen-Image-2.1 INT8 takes ~45 s for 1344×768 on my Spark — while H3 renders video on the same box. Slow per image, but it's the only box where both stay resident.
Totally! At the beginning I thought I only needed one spark. But then I unlocked two. It’s so much fun to see what they can do!! Now trying to accumulate more money for more. The building and learning process itself is much joy for me, which is not that significant for cloud models
@SeeGen_Official Formation changes hold up better than I expected — hands drift a bit in the chorus. I’ve been doing K-pop stage swaps with open models locally; multi-keyframe is exactly what those formation changes need.
@hasancreates@higgsfield_ai Malfoid’s face when the sweater lands 😭 the voices really sell it. Did you do the voice acting in Higgsfield too, or a separate pass?
@FusilladeStudio Wow, truly impressive job!! please receive my respect! hand work is so much under appreciated. Must have gone through so much fine-tuning