Skip to content
Hot eventWatching

是石科技 released a multimodal model service, achieving a 2.48 times speedup in AI video generation.

2 reports2 sourcesUpdated 3 days ago
Subscribe to this event via RSS

Start here

AI overview

Sishi Technology released a multimodal model service, claiming that its video generation inference achieves an average 2.48x speedup end-to-end while maintaining generation quality. Sishi Technology stated that the service uses modality-adaptive token compression, achieving a 1.33x speedup in the WorldSense benchmark; the Context Engine based on the omni-modal model converts material understanding into structured information, supporting the video generation pipeline. The service also incorporates SGLang to optimize cache and GPU memory management.

Written by AI from the reports · updated 3 days ago

Report timeline

Follow the reports to see the event from other sides.

Nov 16
  1. AITNT · AI创作资讯
    Shishi Technology's new solution for AI video 'cost reduction and efficiency enhancement': 2.48x speedup without compromising image quality

    Shishi Technology releases a multimodal model service that employs adaptive acceleration technology based on historical cache residuals, achieving an end-to-end average 2.48x speedup while maintaining generation quality. The service also incorporates SGLang-optimized caching and memory management, and modality-adaptive token compression achieves a 1.33x speedup in the WorldSense benchmark, while the Context Engine bridges material understanding and video generation. The service covers API access, task orchestration, model routing, inference execution, and result delivery, providing support for large-scale AI video production.

  2. 量子位
    Sishi Technology released a multimodal model service, achieving a 2.48x speedup in AI video generation without compromising quality.

    Sishi Technology released a multimodal model service, with video generation inference achieving an average 2.48x speedup end-to-end while maintaining generation quality. The service also uses modality-adaptive token compression, achieving a 1.33x speedup in the WorldSense benchmark. The Context Engine based on the omni-modal model converts material understanding into structured information, supporting the video generation pipeline.

Heat trend for this event

Current
3
Peak
18Nov 16 17:00
24h change
-73%

The trend compares only sources observed the whole time, so the range may be smaller than the current heat figure. Hover or tap the chart for each hour. Left and right arrow keys move through it.