Unify demo outputs and document AudioMAE demo

README.md

@@ -103,22 +103,17 @@ class-agnostic Audio2Sph checkpoint:
 
 ```bash
 said demo --model passt
-```
-
-This command produces `said_demo_1.mp4` through `said_demo_4.mp4` in
-`said_demo_output/`. Each 20-second video presents Ground Truth and Prediction
-side by side over a synchronized 360-degree scene. The four licensed demo
-inputs are included in the release.
-
-The corresponding class-agnostic comparisons use the same four scenes and
-show the merged acoustic field without class labels:
-
-```bash
+said demo --model audiomae
 said demo --model audio2sph
 ```
 
-This command writes `audio2sph_demo_1.mp4` through
-`audio2sph_demo_4.mp4` in `audio2sph_demo_output/`.
+All three commands write to `said_demo_output/`. Their files are named
+`said_passt_demo_1.mp4` through `said_passt_demo_4.mp4`,
+`said_audiomae_demo_1.mp4` through `said_audiomae_demo_4.mp4`, and
+`audio2sph_demo_1.mp4` through `audio2sph_demo_4.mp4`. Each 20-second video
+presents Ground Truth and Prediction side by side over a synchronized
+360-degree scene. The Audio2Sph videos show the merged acoustic field without
+class labels. The four licensed demo inputs are included in the release.
 
 ### Run inference
 

docs/inference.md

@@ -17,7 +17,7 @@ Generate the class-agnostic Audio2Sph comparisons with:
 said demo --model audio2sph
 ```
 
-This writes `audio2sph_demo_output/audio2sph_demo_1.mp4` through
+This writes `said_demo_output/audio2sph_demo_1.mp4` through
 `audio2sph_demo_4.mp4`. Each video merges the official per-source Ground Truth
 maps by a pixelwise maximum on the left and shows the Audio2Sph + Panoramic
 Decoder prediction on the right. Both sides use one fixed zero-to-one display
@@ -28,20 +28,23 @@ The command writes four synchronized comparison videos:
 
 ```text
 said_demo_output/
-├── said_demo_1.mp4
-├── said_demo_2.mp4
-├── said_demo_3.mp4
-└── said_demo_4.mp4
+├── said_passt_demo_1.mp4 ... said_passt_demo_4.mp4
+├── said_audiomae_demo_1.mp4 ... said_audiomae_demo_4.mp4
+└── audio2sph_demo_1.mp4 ... audio2sph_demo_4.mp4
 ```
 
+Each command adds only its four model-specific files. Existing demos from the
+other models remain in the same directory, and an existing file is never
+overwritten.
+
 The generated files use these fixed excerpts:
 
 | Video | Domain | Recording | Interval |
 |---|---|---|---:|
-| `said_demo_1.mp4` | TAU | `fold4_room8_mix003` | 150--170 s |
-| `said_demo_2.mp4` | TAU | `fold4_room10_mix007` | 50--70 s |
-| `said_demo_3.mp4` | Sony | `fold4_room24_mix002` | 20--40 s |
-| `said_demo_4.mp4` | Sony | `fold4_room23_mix003` | 0--20 s |
+| `*_demo_1.mp4` | TAU | `fold4_room8_mix003` | 150--170 s |
+| `*_demo_2.mp4` | TAU | `fold4_room10_mix007` | 50--70 s |
+| `*_demo_3.mp4` | Sony | `fold4_room24_mix002` | 20--40 s |
+| `*_demo_4.mp4` | Sony | `fold4_room23_mix003` | 0--20 s |
 
 Each file is 20 seconds, 1920×600, H.264 video with 48 kHz stereo AAC audio.
 SAID inference operates on the four selected Eigenmike capsules. The