vc405: manifest adds optional full SDXL VAE (vae_decoder_cpu.mnn) for the sharp/fastVae=off decode path

manifest.json

@@ -4,14 +4,14 @@
   "variant": "fp16-encoders + TAESDXL",
   "license": "Stability AI Community License (SDXL-Turbo)",
   "license_note": "Commercial use permitted for orgs/individuals with < $1M USD annual revenue; above that requires a Stability AI Enterprise license. Attribution + license text must be retained with the model. NOT OpenRAIL (that was SD1.5).",
-  "generated": "2026-06-05T03:08:58Z",
+  "generated": "2026-06-05T00:00:00Z",
   "repo": "darkmaniac7/TokForge-SDXL-QNN-NPU (PRIVATE)",
   "notes": [
     "App reads the device Hexagon arch (dsp_arch) and picks the matching set. SDXL ships ONE own-built set today: V75-native (ours).",
     "Forward-compat: a lower-arch bin runs on a higher-arch DSP. The V75-native SDXL set is forward-compat to V79/V81 silicon (verified on the V81 SM8850 handset, .82 + .80). NATIVE V81/V79 SDXL bins are NOT separately built yet.",
     "source=ours: our own clean-room SDXL-Turbo -> W8A16 -> QAIRT 2.40 conversion, drop-in for our libsdxl_qnn_driver (graph name 'model', host combined-emb input). There is no AI-Hub SDXL set in this repo.",
     "verified=true means run on real silicon and confirmed coherent+crisp. verified=false = produced + offline-validated but NEEDS a handset before NPU-enabling.",
-    "VAE path: TAESDXL CPU tiny-VAE (taesdxl_decoder.mnn, 2.4 MB) via --vae_mnn / SDXL_VAE_BACKEND=cpu. The DSP VAE is blocked at 1024px on V75 (unsigned-PD HTP cap) and the OpenCL/Vulkan VAE SIGSEGVs Adreno in Session::resize. CPU TAESDXL is the ONLY validated VAE path.",
+    "VAE: \"Faster image decoding\" ON (default) = TAESDXL CPU tiny-VAE (taesdxl_decoder.mnn, 2.4 MB, fast ~6s, soft) via --vae_mnn / SDXL_VAE_BACKEND=cpu. OFF = full SDXL CPU VAE (vae_decoder_cpu.mnn, 99 MB, cos 1.0 vs torch, sharp ~40s on CPU), also via --vae_mnn (the driver loads either MNN VAE). The DSP VAE is blocked at 1024px on V75 (unsigned-PD HTP cap) and the OpenCL/Vulkan VAE SIGSEGVs Adreno in Session::resize, so BOTH VAEs are CPU MNN. The full VAE is OPTIONAL (optional_files) and downloaded WITH the bundle so it is present when the user turns Faster-decoding off; absence falls back to TAESDXL.",
     "On-device fresh download is ~4.3 GB (4,308,394,979 bytes): unet.bin 2.6 GB + text_encoder_2_fp16 1.4 GB + text_encoder_1_fp16 235 MB + sdxl_emb_mlp 29 MB + taesdxl 2.4 MB + tokenizers/config."
   ],
   "arch_to_socs": {
@@ -143,6 +143,11 @@
           "path": "v75/config.json",
           "size": 936,
           "md5": "833a8f609063eee89c093754af978210"
+        },
+        "vae_decoder_cpu.mnn": {
+          "path": "v75/vae_decoder_cpu.mnn",
+          "size": 99103692,
+          "md5": "5c23735f51d7f9348c351d6224cac16e"
         }
       }
     }
@@ -166,6 +171,9 @@
   ],
   "bundle_dir_name": "SDXL-QNN-NPU",
   "model_id": "image-sdxl-qnn-npu",
-  "total_bytes": 4308394979,
-  "total_human": "4.31 GB (4.01 GiB)"
-}
+  "total_bytes": 4407498671,
+  "total_human": "4.41 GB (4.10 GiB) incl. optional full VAE",
+  "optional_files": [
+    "vae_decoder_cpu.mnn"
+  ]
+}