FlashVSR v1.1 delivered the most overall detail enhancement and our favorite quality/speed balance. If speed is your priority, NVIDIA VSR was the quickest AI option in our RTX 4090 test.
We compared three open-weight approaches—FlashVSR, SeedVR2 and the Community Latent MiniMax Upscale—alongside NVIDIA's SDK and FFmpeg Lanczos. The task: upscale a 768p MiniMax H3 video while preserving its full picture and audio.
Here, 2K means 1440p: 2520×1440, preserving our source's aspect ratio.
All five options
| Upscaler | Why consider it? |
|---|---|
| FlashVSR v1.1 | The strongest overall detail enhancement in our review. Works with a finished video. |
| SeedVR2 7B | Softer, more natural detail enhancement, but about 3.5× slower than FlashVSR in our 4090 test. The extra processing time was not worth it to us. |
| Community Latent MiniMax Upscale | A community model that enlarges H3's saved latents, then refines the result with H3. Runs locally and requires more than an MP4. |
| NVIDIA RTX VSR ULTRA | The fastest AI option in our 4090 test. Uses NVIDIA's SDK. |
| FFmpeg Lanczos | Fastest overall in that test. A conventional CPU resize, without AI-generated detail. |
Compare the videos
Choose a comparison group, then two methods to play together. Community Latent MiniMax Upscale is in the 5090 group; each group uses its own matching source clip. Use 100% detail for faces, fur and edges; watch playback for flicker and unstable textures.
Community Latent MiniMax Upscale uses learned latent upscaling plus refinement. Both outputs below share a source; the four-upscaler comparison uses a different generation.
Shared 768p source
Same source for both upscalersLoading comparison videos…
Pause to compare the same frame. In detail view, drag either panel to pan both. Audio plays once.
Open any original video
Four video upscalers · RTX 4090
Community Latent MiniMax Upscale vs. FlashVSR · RTX 5090
How long does each take?
These are complete processing times for a 6.6-second clip, including model loading from disk into GPU memory and video export. Source-video generation and model downloads are excluded.
On the RTX 4090, Lanczos finished in 29 seconds, NVIDIA VSR in 40 seconds, FlashVSR in 3 minutes 6 seconds, and SeedVR2 in 10 minutes 48 seconds.
On the RTX 5090, the Community Latent MiniMax Upscale with refinement took 7 minutes 54 seconds. FlashVSR processed the matching video in 1 minute 56 seconds—about 4.1× faster. H3 used less peak GPU memory: 20.2 GiB, versus FlashVSR's 26.8 GiB. The H3 visual comparison is still under review.
What does a 5090 buy you?
In our separate FlashVSR hardware comparison, the tuned 5090 had a median time of 1 minute 45 seconds, versus 3 minutes 6 seconds on the 4090: about 43% less waiting.
The Community Latent MiniMax Upscale method
This community method applies a learned upscaler directly to saved H3 latents, then refines the result with H3. It runs locally, without a third-party upscale API.
It needs saved latents and conditioning—the generator's intermediate data. Retain those files and you can run upscaling as a separate job. The other four methods can start from a finished video.
We tested H3's learned upscaler with refinement. The version that skips refinement remains untested.
Our recommendation is FlashVSR v1.1 for the quality/speed balance we preferred. These results cover one short scene at 1440p; they do not establish the best choice for every video or for 4K.
