Update README.md

README.md

@@ -8,6 +8,37 @@ tags:
   - video-enhancement
   - 2x-upscaling
   - safetensors
+widget:
+  - text: "All methods: synchronized 3 x 3 overview"
+    output:
+      url: assets/benchmark/comparison_grid.mp4
+  - text: "Input: 1024 x 768, 243 frames, 24 fps"
+    output:
+      url: assets/benchmark/previews/input.mp4
+  - text: "Case 1: SeedVR2-3B, 2x, 352.820 s, 0.69 inference FPS"
+    output:
+      url: assets/benchmark/previews/seedvr2_3b_x2.mp4
+  - text: "Case 2: FlashVSR+ Turbo, 2x, 54.024 s, 4.50 inference FPS"
+    output:
+      url: assets/benchmark/previews/flashvsr_plus_turbo_x2.mp4
+  - text: "Case 3: Open-H3-Regenerate-2K, 2x, 784.954 s, 0.31 inference FPS"
+    output:
+      url: assets/benchmark/previews/open_h3_regenerate_x2.mp4
+  - text: "Case 4: SwiftVR, 2x, 7.252 s, 33.51 inference FPS"
+    output:
+      url: assets/benchmark/previews/swiftvr_x2.mp4
+  - text: "Case 5: SwiftVR (LightX2V), 2x, 6.399 s, 37.98 inference FPS"
+    output:
+      url: assets/benchmark/previews/swiftvr_lightx2v_x2.mp4
+  - text: "Case 6: RelayVSR, 4x, 26.332 s, 9.23 inference FPS"
+    output:
+      url: assets/benchmark/previews/relayvsr_x4.mp4
+  - text: "Case 7: RelayVSR X4 (LightX2V), 4x, 19.304 s, 12.59 inference FPS"
+    output:
+      url: assets/benchmark/previews/relayvsr_x4_lightx2v.mp4
+  - text: "Case 8: RelayVSR X2 (LightX2V), 2x, 5.348 s, 45.44 inference FPS"
+    output:
+      url: assets/benchmark/previews/relayvsr_x2_lightx2v.mp4
 ---
 
 # RelayVSR_X2
@@ -38,6 +69,43 @@ For inference instructions, see the
 | [Original RelayVSR](https://huggingface.co/kopper/RelayVSR) | 4× | 4W × 4H |
 | [RelayVSR X2](https://huggingface.co/lightx2v/RelayVSR_X2) | 2× | 2W × 2H |
 
+## Inference Benchmark
+
+On the tested clip, **RelayVSR X2 with LightX2V processes 243 frames in 5.348 seconds
+(45.44 inference FPS)** on a single NVIDIA H100 80GB.
+
+Measured on **2026-11-02** using the same
+[drone rooftop input video](assets/benchmark/originals/input.mp4):
+**1024 × 768**, **243 frames**, **24 fps**, **10.125 seconds**.
+All cases ran sequentially on one **NVIDIA H100 80GB HBM3** with
+**PyTorch 2.11.0 + CUDA 13.0**. Each row reports **one measured run after
+same-resolution warmup**.
+
+| Case | Model / implementation | Scale | Native output resolution | Inference time (s) ↓ | Inference FPS ↑ | Original output |
+| --- | --- | --- | --- | ---: | ---: | --- |
+| 1 | SeedVR2-3B | 2× | 2048 × 1536 | 352.820 | 0.69 | [MP4](assets/benchmark/originals/seedvr2_3b_x2.mp4) |
+| 2 | FlashVSR+ Turbo | 2× | 2048 × 1536 | 54.024 | 4.50 | [MP4](assets/benchmark/originals/flashvsr_plus_turbo_x2.mp4) |
+| 3 | Open-H3-Regenerate-2K | 2× | 2048 × 1536 | 784.954 | 0.31 | [MP4](assets/benchmark/originals/open_h3_regenerate_x2.mp4) |
+| 4 | SwiftVR | 2× | 2048 × 1536 | 7.252 | 33.51 | [MP4](assets/benchmark/originals/swiftvr_x2.mp4) |
+| 5 | SwiftVR (LightX2V) | 2× | 2048 × 1536 | 6.399 | 37.98 | [MP4](assets/benchmark/originals/swiftvr_lightx2v_x2.mp4) |
+| 6 | RelayVSR | 4× | 4096 × 3072 | 26.332 | 9.23 | [MP4](assets/benchmark/originals/relayvsr_x4.mp4) |
+| 7 | RelayVSR X4 (LightX2V) | 4× | 4096 × 3072 | 19.304 | 12.59 | [MP4](assets/benchmark/originals/relayvsr_x4_lightx2v.mp4) |
+| 8 | **RelayVSR X2 (LightX2V)** | 2× | 2048 × 1536 | **5.348** | **45.44** | [MP4](assets/benchmark/originals/relayvsr_x2_lightx2v.mp4) |
+
+## Video Comparison
+
+The gallery contains a **synchronized 3 × 3 overview**, the input, and all eight
+restored videos. Individual preview labels follow the benchmark's case numbers.
+The overview layout is:
+
+| Left | Center | Right |
+| --- | --- | --- |
+| Input | SeedVR2-3B · 2× | FlashVSR+ Turbo · 2× |
+| Open-H3 · 2× | SwiftVR · 2× | SwiftVR / LightX2V · 2× |
+| RelayVSR · 4× | RelayVSR X4 / LightX2V · 4× | RelayVSR X2 / LightX2V · 2× |
+
+<Gallery />
+
 ## Model Files
 
 Download both checkpoints into the same model directory: