From 30d19f263b9150ca22e9e7179d6bd5773da1296a Mon Sep 17 00:00:00 2001
From: Sirshendu Ganguly <sirshendu.ganguly2015@gmail.com>
Date: Fri, 3 Apr 2026 21:27:13 +0530
Subject: [PATCH] Reverted to original version of mediaUpload

---
 modules/mediaUpload.py            | 17 +++++++++++++----
 modules/utils/runwareUtils.py     | 30 ------------------------------
 modules/videoInferenceElements.py | 23 +++++++++++++----------
 3 files changed, 26 insertions(+), 44 deletions(-)

diff --git a/modules/mediaUpload.py b/modules/mediaUpload.py
index 2430bd2..1b4f691 100644
--- a/modules/mediaUpload.py
+++ b/modules/mediaUpload.py
@@ -14,7 +14,6 @@ from .utils.runwareUtils import (
     sanitize_for_logging,
     safe_json_dumps,
     sendMediaUUID,
-    convertVideoInputToDataUri,
 )
 
 
@@ -107,9 +106,19 @@ class RunwareMediaUpload:
     def _convertVideoToBase64(self, videoTensor):
         """Convert video file to base64 data URI"""
         print(f"[Debug] VideoFromFile object detected")
-        out = convertVideoInputToDataUri(videoTensor)
-        print(f"[Debug] Processing video file (data URI length {len(out) if out else 0})")
-        return out
+
+        videoFile = self._getVideoFilePath(videoTensor)
+        videoPath = self._extractVideoPath(videoFile)
+
+        print(f"[Debug] Processing video file: {videoPath}")
+
+        with open(videoPath, 'rb') as f:
+            videoBytes = f.read()
+
+        mimeType = self._getMimeType(videoPath)
+        videoBase64 = base64.b64encode(videoBytes).decode('utf-8')
+
+        return f"data:{mimeType};base64,{videoBase64}"
 
     def _getVideoFilePath(self, videoTensor):
         """Extract video file path from VideoFromFile object"""
diff --git a/modules/utils/runwareUtils.py b/modules/utils/runwareUtils.py
index 2dd6f02..0c97b0f 100644
--- a/modules/utils/runwareUtils.py
+++ b/modules/utils/runwareUtils.py
@@ -516,36 +516,6 @@ def convertTensor2IMG(tensorImage):
         return imgDataUri
 
 
-def convertVideoInputToDataUri(video_tensor):
-    """Convert ComfyUI VIDEO (file-backed, e.g. Load Video) to data:<mime>;base64,... for API payloads."""
-    if video_tensor is None:
-        return None
-    video_file = getattr(video_tensor, "_VideoFromFile__file", None)
-    if video_file is None:
-        for attr in ("video_path", "path", "file_path", "filename", "_file"):
-            if hasattr(video_tensor, attr):
-                video_file = getattr(video_tensor, attr)
-                break
-    if video_file is None:
-        raise ValueError(
-            "Could not resolve video file from VIDEO input. Connect a Load Video (file) output."
-        )
-    path = video_file.name if hasattr(video_file, "name") else video_file
-    if not isinstance(path, str):
-        raise ValueError(f"Unexpected video file type: {type(video_file)}")
-    with open(path, "rb") as f:
-        video_bytes = f.read()
-    ext = os.path.splitext(path.lower())[1]
-    mime = {
-        ".mp4": "video/mp4",
-        ".avi": "video/avi",
-        ".mov": "video/quicktime",
-        ".webm": "video/webm",
-    }.get(ext, "video/mp4")
-    b64 = base64.b64encode(video_bytes).decode("utf-8")
-    return f"data:{mime};base64,{b64}"
-
-
 def convertTensor2IMGBase64Only(tensorImage):
     """Convert tensor to base64 data URI without caching - for frame images"""
     imageNP = (tensorImage.squeeze().numpy() * 255).astype(np.uint8)
diff --git a/modules/videoInferenceElements.py b/modules/videoInferenceElements.py
index e10e12f..f1f0d70 100644
--- a/modules/videoInferenceElements.py
+++ b/modules/videoInferenceElements.py
@@ -50,7 +50,7 @@ def _build_element_dict(
     description: str,
     frontalImage: Any,
     refer_images: List[Any],
-    refer_videos: List[Any],
+    refer_videos: List[str],
     voice: str,
     tags: str,
 ) -> Dict[str, Any]:
@@ -70,9 +70,8 @@ def _build_element_dict(
     imgs = [x for x in refer_images if x is not None]
     if imgs:
         el["images"] = imgs
-    vids = [x for x in refer_videos if x is not None]
-    if vids:
-        el["videos"] = vids
+    if refer_videos:
+        el["videos"] = refer_videos
     voices = _split_lines_or_csv(voice)
     if voices:
         el["voice"] = voices
@@ -116,8 +115,12 @@ class RunwareVideoInferenceElements:
                 "tooltip": f"Optional reference image {i} for image_refer (up to {_REF_IMAGE_SLOTS}).",
             })
         for i in range(1, _REF_VIDEO_SLOTS + 1):
-            optional[f"videos_{i}"] = ("VIDEO", {
-                "tooltip": f"Optional reference video {i} for video_refer (file-backed VIDEO, e.g. Load Video).",
+            optional[f"videos_{i}"] = ("STRING", {
+                "default": "",
+                "tooltip": (
+                    f"Optional reference video {i}: public URL or mediaUUID from Runware Media Upload "
+                    f"(upload a video first, then paste the UUID here)."
+                ),
             })
         return {"required": {}, "optional": optional}
 
@@ -143,11 +146,11 @@ class RunwareVideoInferenceElements:
             if slot is not None:
                 refer_images.append(rwUtils.convertTensor2IMG(slot))
 
-        refer_videos: List[Any] = []
+        refer_videos: List[str] = []
         for i in range(1, _REF_VIDEO_SLOTS + 1):
-            slot = kwargs.get(f"videos_{i}")
-            if slot is not None:
-                refer_videos.append(rwUtils.convertVideoInputToDataUri(slot))
+            raw = kwargs.get(f"videos_{i}")
+            if raw is not None and str(raw).strip() != "":
+                refer_videos.append(str(raw).strip())
 
         el = _build_element_dict(
             str(kwargs.get("id") or ""),