Document measured runtime and the soundfile WAV-write delta

README.md

@@ -34,7 +34,8 @@ so the picture and the soundtrack are denoised together rather than dubbed after
 ## What it produces
 
 The checkpoint's native shape: **121 frames · 256×256 · 24 fps · 48 kHz audio (~5 s)**,
-muxed into a single MP4. A run takes a few minutes on ZeroGPU.
+muxed into a single MP4. A default 50-step run takes ~97 s on ZeroGPU (measured:
+~1.9 s/step plus ~3 s for text encoding, VAE decoding and muxing).
 
 ## Faithfulness to the reference implementation
 
@@ -65,6 +66,9 @@ Deployment-only deltas from the upstream tree, all of which leave the numerics u
   optional.
 - `joint_av/validation.py` — the denoising loop is wrapped in `tqdm` so the UI can show
   progress.
+- `torchaudio.save` is routed through `soundfile` in `app.py`: torchaudio ≥ 2.10 delegates
+  `save` to `torchcodec`, whose wheels are bound to a specific FFmpeg ABI. Both write the
+  same 48 kHz PCM WAV.
 - Components are loaded on CPU and then moved with `.to("cuda")`; upstream's
   `device="cuda"` path uses `device_map={"": "cuda:0"}` + `torch.cuda.set_device`, which
   ZeroGPU does not support.