Unverified50% confidenceFactExact time
When using the 0.8B model, the first call requires loading the model and takes noticeably longer; subsequent calls are extremely fast, and the total time for one complete call in the video is about 0.23 seconds.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/27/2026
First Seen
Valid until: 11/10/2026
Sources
Related Entities
Related Claims
UnverifiedVideo real-time processing typically requires completing single-frame processing within 33 milliseconds, whereas VLM inference latency far exceeds this requirement, and a pure VLM solution may be 10 to 100 times slower than real-time.77% similarVerifiedWith the accelerator enabled, MiniMax H3 takes approximately 2 minutes to generate 1 second of video, approximately 14 minutes for a 480P ten-second video, and approximately 20–30 minutes for a 768P ten-second video.76% similarUnverifiedAfter setting the VRAM headroom, the standard deviation of per-segment video generation time dropped from 1.62 to 0.11, and the worst-case outliers fell from 21–25 seconds to 15.73 seconds.76% similarUnverifiedOn the same device, simply removing the 30fps frame rate cap during the preloading phase reduced load time from 4.1 seconds to 1.7 seconds.76% similarUnverifiedThe author states that this long-video workflow is not intended to replace the existing 30-second solution, as generating 60-second videos on a 12 GB GPU is more time-consuming.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1104420API
curl https://kongchang.com/api/v1/knowledge/claims/1104420MCP
get_claim(id=1104420)