Measured RTX 4070 timings

README.md

@@ -28,7 +28,7 @@ second half of each clip to get smeary and melty.** That's the price of half the
 | Denoiser | 722,386,240 params (bf16, 1.44 GB): 15 of the teacher's 30 Wan blocks (0,2,4,…,12,15,17,…,29) |
 | Also needs | Wan2.1 VAE (0.5 GB) + umT5-XXL text encoder (11.4 GB), downloaded automatically from `Wan-AI/Wan2.1-T2V-1.3B` |
 | Output | 480×832, 33 frames @ 16 fps (up to 81 frames possible, meltier) |
-| Speed | ~2 s on an H100, ~10 s on an L4, ~25 s on a T4 (plus text encoding for a new prompt) |
+| Speed | ~2 s on an H100, ~10 s on an L4 or RTX 4070, ~25 s on a T4 (plus text encoding for a new prompt) |
 | GPU memory | 5.6 GB peak with the text encoder on CPU; runs on 8–12 GB cards |
 
 ## Quick start
@@ -48,7 +48,7 @@ Command line: `python fuku40.py --image my_photo.jpg --prompt "..." --out out.mp
 
 - **Text encoder modes** (`text_mode=` / `--text-mode`, picked automatically from free VRAM):
   `gpu` (40+ GB cards), `reload` (16 GB cards: Colab/Kaggle T4; loads per new prompt),
-  `cpu` (≤12 GB cards, ~30–90 s per new prompt on the CPU; generation itself stays on the GPU).
+  `cpu` (≤12 GB cards, ~1–2 min per new prompt on the CPU; generation itself stays on the GPU).
 - Images are center-cropped to landscape 832×480, the only size it was trained on.
 - Set `HF_HOME` / `FUKU40_HOME` to put the downloads on a drive with space.