Initial commit: ComfyUI-VideoToText

2026-05-22 09:33:52 +08:00
commit 7e0e1bf4fb
21 changed files with 1964 additions and 0 deletions
@@ -0,0 +1,71 @@
+{
+  "loader_video": {
+    "class_type": "VTT_LoadVideo",
+    "inputs": {
+      "video": "example.mp4",
+      "video_path_override": ""
+    }
+  },
+  "extract_audio": {
+    "class_type": "VTT_ExtractAudio",
+    "inputs": {
+      "video_path": ["loader_video", 0],
+      "sample_rate": 16000,
+      "mono": true
+    }
+  },
+  "audio_in": {
+    "class_type": "VTT_AudioFromPath",
+    "inputs": {
+      "audio_path": ["extract_audio", 0],
+      "target_sample_rate": 0,
+      "force_mono": false
+    }
+  },
+  "mimo_loader": {
+    "class_type": "MiMoASRLoader",
+    "inputs": {
+      "model_path": "/big_disk/models/MiMo-V2.5-ASR",
+      "tokenizer_path": "/big_disk/models/MiMo-Audio-Tokenizer",
+      "repo_path": "",
+      "force_reload": false
+    }
+  },
+  "mimo_asr": {
+    "class_type": "MiMoASRTranscribe",
+    "inputs": {
+      "model": ["mimo_loader", 0],
+      "audio": ["audio_in", 0],
+      "language": "Auto"
+    }
+  },
+  "build_json": {
+    "class_type": "VTT_BuildTranscriptJSON",
+    "inputs": {
+      "text": ["mimo_asr", 0],
+      "audio_path": ["extract_audio", 0],
+      "duration_sec": ["audio_in", 1],
+      "model_name": "MiMo-V2.5-ASR",
+      "tokenizer_name": "MiMo-Audio-Tokenizer",
+      "language": "Auto",
+      "source_video_path": ["loader_video", 0]
+    }
+  },
+  "save": {
+    "class_type": "VTT_SaveTranscript",
+    "inputs": {
+      "transcript_json": ["build_json", 0],
+      "filename_prefix": "transcript",
+      "include_timestamp": true,
+      "subfolder": ""
+    }
+  },
+  "preview_text": {
+    "class_type": "PreviewAny",
+    "inputs": {"source": ["mimo_asr", 0]}
+  },
+  "preview_json": {
+    "class_type": "PreviewAny",
+    "inputs": {"source": ["build_json", 0]}
+  }
+}