Activity, over time
Daily GitHub activity and social mentions.
Mentions
Daily posts linking to this repository
Attention
Daily stars & forks
Star accumulation
Cumulative stars
Pull requests
Daily opens & merges
Conversations
Daily issues & comments
Community
Daily accounts & reviews
Trending history
Daily top 10 · last 90 days · UTC
Mentions
Stars
Forks
Contributors
Latest stargazers
The last 10 distinct accounts we observed this year.
Most active participants
PR, issue, comment & review events in the past year.
Packages
No packages found.
Live mentions badge
Show how many posts mention your project in the last 30 days.
Live mentions · last 30 days
Mentions
Prism uses an adjustable stencil: it shapes spatiotemporal blocks from visual variation and audio-video coupling, then selects blocks per query. Unlike fixed full attention, it can miss relevant regions. Inspect the released config; sparsity isn’t free. Tencent-Hunyuan/Prism
Prism (Fudan + Tencent Hunyuan): image-to-video with the sound baked in. Speech, SFX and music, native 720p up to 2K. Preview weights are MIT. Paper: arxiv.org Weights: huggingface.co Code: Tencent-Hunyuan/Prism t.co
Prism is dynamic sparse attention for training 2K joint video-audio models. Authors report a 2.5x training speedup over full attention, with preview checkpoints. MIT code. Tencent-Hunyuan/Prism