阿里刚刚放出来了一款环境音/音效模型：PrismAudio，5.18亿参数，生成9秒音频只需0.63秒 它把强化学习和思维链结合，先思考，再生成匹配的声音
声画同步、以及音质上的清晰度丰富度感觉还可以
#音效模型 #PrismAudio
00:45
2 Replies
12 Retweets
37 Likes
3,938 Views 
One Sentence Summary
Alibaba launches PrismAudio, a 518M parameter sound effect model supporting audio-visual synchronization and rapid generation.
Summary
Alibaba has released PrismAudio, a sound effect model with 518 million parameters. By combining reinforcement learning with Chain-of-Thought technology, it achieves rapid audio generation (9 seconds of audio in 0.63 seconds) while delivering strong audio-visual synchronization and high-quality sound output.
AI Score
85
Influence Score 18
Published At Today
Language
Chinese
Tags
PrismAudio
Sound Effect Model
Audio Generation
AI Model