10x Faster Real-time High-quality AI Video Generation
Run state-of-the-art video models and generate high quality videos faster on Tenstorrent Galaxy superclusters.
Lower latency, More throughput
Faster than real-time
The industry’s fastest video generation is now 10x faster on Tenstorrent Galaxy supercluster.
Latency (sec/video)
Wan 2.2 5B
Wan 2.2 A14B Lightning, Nvidia x Prodia
grok-imagine-video, xAI
Wan 2.2 A14B, Tenstorrent x Prodia
More high-quality frames per second
Produce more high-quality videos using state-of-the-art models on Tenstorrent Galaxy supercluster.
Throughput (fps)
Wan 2.2 5B
Wan 2.2 A14B Lightning, Nvidia x Prodia
grok-imagine-video, xAI
Wan 2.2 A14B, Tenstorrent x Prodia
Benefits
Fast
10x faster than leading GPU setups at comparable configurations. Our optimized diffusion transformer library accelerates all leading video models.
Quality
Acceleration doesn’t have to come at the cost of quality. 10x faster at the same precision. Tenstorrent Galaxy enables high-quality video with fast inference.
Scalable
Add Tenstorrent Galaxy superclusters to add speed. Larger models, higher resolution. Lower latency, lower cost.
Open
Fully open-source software stack. No proprietary runtime locking your video pipeline to a single vendor.
Technologies
Blackhole Architecture for Real-Time Video Generation
Tenstorrent Galaxy superclusters
Run anything – fast, affordable, simple. High-density, scalable compute. Add systems, add speed.
Tensix Cores
Purpose-built for parallel, continuous workloads. With 91x SRAM/capacity per dollar and 12x SRAM bandwidth per dollar. Tensix delivers where it counts.
Model Support
90% of models from HuggingFace just work and coverage is growing every day across LLMs, Image Gen, Speech, Vision, Embeddings, Encoders and more.
1x Tenstorrent Galaxy™ Blackhole supercluster
Tenstorrent Galaxy™ Blackhole can be deployed in superclusters, extending into multi-server topologies that can scale-out to any size. Tenstorrent Galaxy™ superclusters lead the industry in performance and cost for faster than real-time AI video generation.
