whereismyai
FlashVSR Plus
FlashVSR Plus: FlashVSR+ is an efficient one-step diffusion framework for high-quality video super-resolution.
FlashVSR+ is an efficient one-step diffusion framework designed for high-quality video super-resolution. It enables spatially untiled 4K inference on a single 24GB RTX 4090 GPU through hybrid attention and reward-yield distribution matching distillation. The framework provides three presets (Turbo, Base, and V1) accessible via a command-line interface or a Python API. It supports Linux operating systems, NVIDIA GPUs with CUDA, and Python 3.10. It also offers configurable attention backends, including Block Sparse Attention and FlashVSR Triton Attention, along with fallback options.
FlashVSR Plus at a glance
Key features
- Run Turbo, Base, or V1 presets
- Support large-motion videos
- Process long videos in segments
- Use block sparse or Triton attention backends
- Enable low-frequency color correction
- Execute through Python API or CLI
Use it for
- Enhance video resolution up to 4K
- Process videos with large motion details
- Run high-quality video super-resolution on local hardware
Alternatives to FlashVSR Plus
- TurboDiffusion — Accelerates video diffusion generation by up to 200 times on single GPUs.
- FastVideo — Provides end-to-end post-training and real-time inference for accelerated video generation.
- FreeVideo — Runs MiniMax H3 locally on consumer GPUs with as little as 8 GB of VRAM.
- Self Gradient Forcing Plus — It enables continuous video generation for up to 24 hours from 5-second training windows.
- SolarRenderer — Renders game engine and agent structures into finished video in four steps.
- Kandinsky 6 — Generates synchronized 5-second video clips and 44 kHz audio using diffusion models.
- Bernini — Combines a semantic planner and DiT renderer for video generation and editing tasks.
- Wan — Provides native 30-second video generation and pixel-perfect consistency.
- LightX2V — It combines avatars, image-to-video, and image generation in one workspace.
- H3.c — Runs native MiniMax-H3 video and audio generation locally on Apple Silicon.
- ATAS — Maintains persistent spatial and event memory across shots.
- OneStreamer — It jointly learns to record evidence and respond through proactive generation.
All FlashVSR Plus alternatives