初始提交：3 款纯音频游戏 + 共享引擎 + 完整文档与资产管线

- 共享引擎（app/src/main）：GestureEngine 6 手势识别、AudioEngine 三层音频（ExoPlayer BGM 交叉淡入 / 对白 ducking / SoundPool 低延迟）、Game/GameEngine 状态机接口、MainActivity 全屏封面 + TalkBack - 《声命线》storyline：末世互动广播剧，3 幕 8 抉择点 4 结局 - 《声东击西》reflex：听声辨位反应，12 波战斗 + BGM 三档 BPM 变速 - 《暗室》escape：3x3 网格密室逃脱，钥匙→铁柜→保险丝→电箱→大门谜题链 - 文档：setup（从零搭建）/ architecture（架构）/ engine-api（引擎 API）/ asset-pipeline（资产流水线）/ games/*（三款游戏设计文档） - 资产管线：Qwen3-TTS 对白 + 程序合成音效/BGM + Mureka BGM 的 gen 脚本与台账，全部 ogg 资产入库可复现 - env.d 环境脚本：android/tts 入库；sfx/mureka 含密钥不入库，仅提交 .example.sh 模板
2026-08-29 13:00:34 +08:00
commit f0f0a9b50e
446 changed files with 2627836 additions and 0 deletions
@@ -0,0 +1,116 @@
+from dataclasses import dataclass
+
+from huggingface_hub import hf_hub_download, try_to_load_from_cache
+
+
+@dataclass(frozen=True)
+class ModelConfig:
+    repo_id: str
+    config_path: str
+    ckpt_path: str
+
+    def resolve(self):
+        """Download files from HuggingFace Hub and return local cached paths."""
+        local_config = hf_hub_download(repo_id=self.repo_id, filename=self.config_path)
+        local_ckpt = hf_hub_download(repo_id=self.repo_id, filename=self.ckpt_path)
+        return local_config, local_ckpt
+
+
+@dataclass(frozen=True)
+class AutoencoderModelConfig:
+    """Config for a standalone autoencoder HF repo (e.g. stabilityai/SAME-S).
+
+    resolve() first checks whether any full Stable Audio 3 checkpoint that contains the same
+    autoencoder is already cached locally.  If one is found it is used as-is
+    (load_autoencoder strips the pretransform.* prefix automatically), avoiding a
+    redundant download.  Otherwise the lightweight AE-only repo is fetched instead.
+    """
+
+    ae_repo_id: str
+    ae_config_path: str
+    ae_ckpt_path: str
+    stable_audio_3: tuple[ModelConfig, ...]
+
+    def resolve(self):
+        """Return (config_path, ckpt_path), preferring an already-cached Stable Audio 3 checkpoint."""
+        for fallback in self.stable_audio_3:
+            cached_config = try_to_load_from_cache(
+                fallback.repo_id, fallback.config_path
+            )
+            cached_ckpt = try_to_load_from_cache(fallback.repo_id, fallback.ckpt_path)
+            if isinstance(cached_config, str) and isinstance(cached_ckpt, str):
+                return cached_config, cached_ckpt
+
+        # No Stable Audio 3 checkpoint found in local cache — download the AE-only repo.
+        local_config = hf_hub_download(
+            repo_id=self.ae_repo_id, filename=self.ae_config_path
+        )
+        local_ckpt = hf_hub_download(
+            repo_id=self.ae_repo_id, filename=self.ae_ckpt_path
+        )
+        return local_config, local_ckpt
+
+
+models: dict[str, ModelConfig] = {
+    "small-music": ModelConfig(
+        "stabilityai/stable-audio-3-small-music",
+        "model_config.json",
+        "model.safetensors",
+    ),
+    "small-music-base": ModelConfig(
+        "stabilityai/stable-audio-3-small-music-base",
+        "model_config.json",
+        "model.safetensors",
+    ),
+    "small-sfx": ModelConfig(
+        "stabilityai/stable-audio-3-small-sfx",
+        "model_config.json",
+        "model.safetensors",
+    ),
+    "small-sfx-base": ModelConfig(
+        "stabilityai/stable-audio-3-small-sfx-base",
+        "model_config.json",
+        "model.safetensors",
+    ),
+    "medium": ModelConfig(
+        "stabilityai/stable-audio-3-medium",
+        "model_config.json",
+        "model.safetensors",
+    ),
+    "medium-base": ModelConfig(
+        "stabilityai/stable-audio-3-medium-base",
+        "model_config.json",
+        "model.safetensors",
+    ),
+}
+
+# Stable Audio 3 full-model configs to probe (in order) before downloading the AE-only repo.
+_small_stable_audio_3: tuple[ModelConfig, ...] = (
+    models["small-music"],
+    models["small-sfx"],
+)
+_medium_stable_audio_3: tuple[ModelConfig, ...] = (models["medium"],)
+
+ae_models: dict[str, AutoencoderModelConfig] = {
+    "same-s": AutoencoderModelConfig(
+        ae_repo_id="stabilityai/SAME-S",
+        ae_config_path="model_config.json",
+        ae_ckpt_path="model.safetensors",
+        stable_audio_3=_small_stable_audio_3,
+    ),
+    "same-l": AutoencoderModelConfig(
+        ae_repo_id="stabilityai/SAME-L",
+        ae_config_path="model_config.json",
+        ae_ckpt_path="model.safetensors",
+        stable_audio_3=_medium_stable_audio_3,
+    ),
+}
+
+base_models: dict[str, ModelConfig] = {
+    k: v for k, v in models.items() if k.endswith("-base")
+}
+
+all_models: dict[str, ModelConfig | AutoencoderModelConfig] = {
+    **models,
+    **ae_models,
+}