REAL Video Enhancer
Desktop app for video frame interpolation and upscaling with RIFE, GMFSS and SPAN models on TensorRT, PyTorch or NCNN.
RIFE, GMFSS and upscaling models on TensorRT, PyTorch or NCNN for Windows, Linux and macOS.
REAL Video Enhancer is a Qt desktop app that raises video frame rates and upscales resolution with neural models. It positions itself as an alternative to older tools such as Flowframes, and suits people with an NVIDIA RTX or Vulkan GPU.
What it is
You choose an interpolation model such as RIFE, GMFSS or IFRNet, or an upscaling model such as the SPAN series, AnimeJaNai, AniSD or AnimeSR, then run it on a TensorRT, PyTorch or NCNN backend. It adds scene change detection to keep sharp cuts and a preview of the latest rendered frame, and it also lists decompression and denoise models.
Who it's for
It suits home users enhancing anime, game captures or old footage, who want a GUI rather than a command line. It does not suit low-spec machines: the minimum is 16 GB RAM and a Vulkan 1.3 GPU.
Key features
Frame interpolation
RIFE, GMFSS and IFRNet models double frame rates, for example 24 to 48 FPS.
Video upscaling
SPAN, AnimeJaNai, AniSD and AnimeSR models raise resolution.
Three backends
TensorRT for NVIDIA RTX, PyTorch for CUDA or ROCm, NCNN for Vulkan.
Scene change detection
Keeps sharp transitions from being blended.
Live frame preview
Shows the latest rendered frame during a job.
Decompression and denoise
Optional models such as DeH264, DRUnet and DnCNN.
Up and running in 3 steps
Our honest verdict
What's great
- Packaged for Windows, Ubuntu 22.04+ (executable and Flatpak) and macOS 15+, plus a Steam listing
- NCNN backend runs on a wider range of Vulkan GPUs for people without an RTX card
- Wide model selection across interpolation, upscaling, decompression and denoise
- Free Colab notebook for people without a capable GPU
Where it falls short
- Hardware needs are real: README minimum is 16 GB RAM and a Vulkan 1.3 GPU, and the recommended setup is an NVIDIA RTX card with 8 GB VRAM and 32 GB RAM
- The README warns the Windows binary can be flagged as a trojan, which it says is a false positive from PyInstaller
- TensorRT takes a long first optimisation per video resolution and can fail on limited VRAM; the README calls ROCm buggy
- AGPL-3.0 licence, and each bundled model has its own author and licence that you should check
Pricing & license
Free under AGPL-3.0 and listed as Free to Play on Steam. The README links a Ko-fi page for optional support, and your only other cost is the GPU hardware you run it on.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Do I need an NVIDIA GPU?
Not strictly. NCNN runs on any Vulkan 1.3 GPU, but the README recommends an NVIDIA RTX 20-series or newer for TensorRT.
Is the Steam version paid?
The Steam page lists it as Free to Play.
Why does my antivirus flag it?
The README says the Windows binary can be detected as a trojan and describes this as a false positive caused by PyInstaller.
