Record the measured GPU costs

README.md

@@ -81,6 +81,22 @@ The VAEs move on the first GPU call. `H3_GPU_SIZE=xlarge` is required — `large
 [`multimodalart/minimax-h3-aoti`](https://huggingface.co/multimodalart/minimax-h3-aoti), one ahead-of-time-compiled
 transformer block serving all 50, which removes roughly 0.5 s/step.
 
+## Measured
+
+The default request — 960x544, 124 frames (5.167 s), 15 steps of `linear_quadratic`, RCAS 0.3, FILM 2x to 48 fps:
+
+| | |
+|---|---|
+| startup | 97–153 s (77.3 GB over Xet, 50 AoTI blocks patched) |
+| denoise + both decoders | 61 s, 4.1 s/step |
+| RCAS + FILM + h264 mux | 9 s |
+| **booked ZeroGPU** | **89 s** for a 70 s call |
+
+First-and-last-frame at the same canvas costs 79 s — the two keyframes add 510 conditioning rows each — and its
+output's first and last frames reproduce the supplied keyframes to a mean absolute error of 2–3/255. The estimator is
+quadratic in the canvas, so 1344x768 is roughly 2.7x the denoise of 960x544; every run prints its measured GPU time
+against what it booked.
+
 ## Files
 
 | file | |

app.py

@@ -212,8 +212,10 @@ def encode_remote(prompt, image_path, last_image_path, canvas, num_frames, rewri
         return handle.get_tensor("prompt_embeds"), handle.get_tensor("text_token_tags"), metadata, plan
 
 
-# Seconds of GPU one request needs. Fitted to *this* Space, measured: booking a ceiling nobody reaches spends every
-# visitor's ZeroGPU quota on nothing and costs the demo queue priority, so each term below is a measurement.
+# Seconds of GPU one request needs. Fitted to *this* Space against measurements, because booking a ceiling nobody
+# reaches spends every visitor's ZeroGPU quota on nothing and costs the demo queue priority. Measured on the live
+# Space: the default request takes 70 s and books 89; the first-and-last-frame one takes 79 s and books 94. The report
+# each request prints carries both numbers, so the fit stays checkable.
 #
 # The denoise loop, from the packed video rows it is about to run: linear in the rows for the matmuls, quadratic for
 # the attention, against the AoTI block package this Space loads. 3.6 s/step at the default canvas.