Add batch processing support for faster frame interpolation

Processes multiple frame pairs simultaneously instead of one-by-one.
New batch_size input (1-64) lets users trade VRAM for speed.
Refactored pyr_level logic into shared _get_pyr_level() helper.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

inference.py

@@ -43,6 +43,18 @@ class BiMVFIModel:
         self.model.to(device)
         return self
 
+    def _get_pyr_level(self, h):
+        if self.auto_pyr_level:
+            if h >= 2160:
+                return 7
+            elif h >= 1080:
+                return 6
+            elif h >= 540:
+                return 5
+            else:
+                return 3
+        return self.pyr_level
+
     @torch.no_grad()
     def interpolate_pair(self, frame0, frame1, time_step=0.5):
         """Interpolate a single frame between two input frames.
@@ -59,19 +71,36 @@ class BiMVFIModel:
         img0 = frame0.to(device)
         img1 = frame1.to(device)
 
-        if self.auto_pyr_level:
+        pyr_level = self._get_pyr_level(img0.shape[2])
-            _, _, h, _ = img0.shape
+        time_step_tensor = torch.tensor([time_step], device=device).view(1, 1, 1, 1)
-            if h >= 2160:
+
-                pyr_level = 7
+        result_dict = self.model(
-            elif h >= 1080:
+            img0=img0, img1=img1,
-                pyr_level = 6
+            time_step=time_step_tensor,
-            elif h >= 540:
+            pyr_level=pyr_level,
-                pyr_level = 5
+        )
-            else:
+
-                pyr_level = 3
+        interp = result_dict["imgt_pred"]
-        else:
+        interp = torch.clamp(interp, 0, 1)
-            pyr_level = self.pyr_level
+        return interp
-
+
+    @torch.no_grad()
+    def interpolate_batch(self, frames0, frames1, time_step=0.5):
+        """Interpolate multiple frame pairs at once.
+
+        Args:
+            frames0: [B, C, H, W] tensor, float32, range [0, 1]
+            frames1: [B, C, H, W] tensor, float32, range [0, 1]
+            time_step: float in (0, 1), temporal position of interpolated frames
+
+        Returns:
+            Interpolated frames as [B, C, H, W] tensor, float32, clamped to [0, 1]
+        """
+        device = next(self.model.parameters()).device
+        img0 = frames0.to(device)
+        img1 = frames1.to(device)
+
+        pyr_level = self._get_pyr_level(img0.shape[2])
         time_step_tensor = torch.tensor([time_step], device=device).view(1, 1, 1, 1)
 
         result_dict = self.model(

nodes.py

@@ -119,6 +119,10 @@ class BIMVFIInterpolate:
                     "default": False,
                     "tooltip": "Store all intermediate frames on GPU instead of CPU. Much faster (no transfers) but requires enough VRAM for all frames. Recommended for 48GB+ cards.",
                 }),
+                "batch_size": ("INT", {
+                    "default": 1, "min": 1, "max": 64, "step": 1,
+                    "tooltip": "Number of frame pairs to process simultaneously. Higher = faster but uses more VRAM. Start with 1, increase until VRAM is full. Recommended: 1 for 8GB, 2-4 for 24GB, 4-16 for 48GB+.",
+                }),
             }
         }
 
@@ -127,7 +131,7 @@ class BIMVFIInterpolate:
     FUNCTION = "interpolate"
     CATEGORY = "video/BIM-VFI"
 
-    def interpolate(self, images, model, multiplier, clear_cache_after_n_frames, keep_device, all_on_gpu):
+    def interpolate(self, images, model, multiplier, clear_cache_after_n_frames, keep_device, all_on_gpu, batch_size):
         if images.shape[0] < 2:
             return (images,)
 
@@ -161,26 +165,32 @@ class BIMVFIInterpolate:
             new_frames = []
             num_pairs = frames.shape[0] - 1
 
-            for i in range(num_pairs):
+            for i in range(0, num_pairs, batch_size):
-                frame0 = frames[i:i+1]   # [1, C, H, W]
+                batch_end = min(i + batch_size, num_pairs)
-                frame1 = frames[i+1:i+2] # [1, C, H, W]
+                actual_batch = batch_end - i
+
+                # Gather batch of pairs
+                frames0 = torch.cat([frames[j:j+1] for j in range(i, batch_end)], dim=0)
+                frames1 = torch.cat([frames[j+1:j+2] for j in range(i, batch_end)], dim=0)
 
                 if not keep_device:
                     model.to(device)
 
-                mid = model.interpolate_pair(frame0, frame1, time_step=0.5)
+                mids = model.interpolate_batch(frames0, frames1, time_step=0.5)
-                mid = mid.to(storage_device)
+                mids = mids.to(storage_device)
 
                 if not keep_device:
                     model.to("cpu")
 
-                new_frames.append(frames[i:i+1])
+                # Interleave: original frame, then interpolated frame
-                new_frames.append(mid)
+                for j in range(actual_batch):
+                    new_frames.append(frames[i + j:i + j + 1])
+                    new_frames.append(mids[j:j+1])
 
-                step += 1
+                step += actual_batch
                 pbar.update_absolute(step, total_steps)
 
-                if not all_on_gpu and (i + 1) % clear_cache_after_n_frames == 0 and torch.cuda.is_available():
+                if not all_on_gpu and (batch_end) % clear_cache_after_n_frames == 0 and torch.cuda.is_available():
                     clear_backwarp_cache()
                     torch.cuda.empty_cache()