Avataar、インド市場向けに動画生成コストを20分の1に削減
Avataarは動画モデル「varya」を公開し、1秒あたり0.005ドルでクリップを生成。Runwayなどの競合より20倍安い。この価格設定は、価格に敏感で動画ファーストのインドのコンシューマー市場を狙ったものだ。
backed by peak xv and the india ai mission, avataar.ai released varya, a video generation model distilled from alibaba's wan 2.2. the model runs in 4 steps compared to 50, producing video 10 times faster on a single h200 gpu and at a cost of $0.005 per second. it is trained on curated data to recognize indian cultural nuances in food, clothing, and festivals, and will be released with open weights on the government's ai kosh portal. this move highlights india's pragmatic approach of prioritizing cost-effective applications over expensive foundation model training to drive mass adoption.