From 78bb8d5665e60be2d65aa76ae0a95f78ff862b4c Mon Sep 17 00:00:00 2001
From: Meisoftcoltd <38667978+Meisoftcoltd@users.noreply.github.com>
Date: Mon, 6 Apr 2026 08:45:00 +0000
Subject: [PATCH] fix: resolve race condition in IncrementalVideoStitcher

Refactored `IncrementalVideoStitcher` to evaluate the sequence termination condition autonomously.
Replaced reliance on a potentially desynced global frame accumulator with a class-level variable `_autonomous_frame_count` that natively increments based on incoming tensor batch sizes. This ensures deterministic behavior regardless of ComfyUI's internal node execution order, eliminating premature video completions on single-cycle generations.
---
 video.py | 21 ++++++++++++++++++---
 1 file changed, 18 insertions(+), 3 deletions(-)

diff --git a/video.py b/video.py
index b6c857a..ca0ae5d 100644
--- a/video.py
+++ b/video.py
@@ -296,6 +296,9 @@ class AutoLoopCalculator:
 
 @register_node
 class IncrementalVideoStitcher:
+    # 1. Variable de clase para memoria interna autónoma
+    _autonomous_frame_count = 0
+
     @classmethod
     def INPUT_TYPES(s):
         return {
@@ -315,6 +318,13 @@ class IncrementalVideoStitcher:
         from . import loop
         import os, folder_paths, torch
 
+        # 2. Resetear el contador si es un renderizado nuevo
+        if current_loop_index == 0:
+            IncrementalVideoStitcher._autonomous_frame_count = 0
+
+        # 3. Sumar inmediatamente los frames que acabamos de recibir
+        IncrementalVideoStitcher._autonomous_frame_count += images.shape[0]
+
         cache_dir = os.path.join(folder_paths.get_temp_directory(), "wan_stitcher_cache")
         os.makedirs(cache_dir, exist_ok=True)
 
@@ -322,11 +332,16 @@ class IncrementalVideoStitcher:
         torch.save(images.cpu(), path)
         print(f"🎞️ [Stitcher] Lote {current_loop_index} guardado en disco ({images.shape[0]} frames).")
 
+        # 4. Calcular el objetivo matemático
         source_total = getattr(loop, 'global_source_frame_count', 1)
-        is_final_cycle = loop.global_accumulated_frames >= source_total
+        nth = getattr(loop, 'global_select_every_nth', 1)
+        expected_total_frames = source_total // nth
+
+        # 5. Evaluación 100% ciega a las variables globales de progreso
+        is_final_cycle = IncrementalVideoStitcher._autonomous_frame_count >= expected_total_frames
 
         if is_final_cycle:
-            print(f"   -> 🏁 Generación completa. Liberando tensor total al flujo para procesos finales.")
+            print(f"   -> 🏁 Generación completa ({IncrementalVideoStitcher._autonomous_frame_count}/{expected_total_frames}). Liberando tensor total al flujo para procesos finales.")
             print(f"📦 [Stitcher] Ensamblando todos los lotes de vídeo...")
             all_tensors = []
             for i in range(current_loop_index + 1):
@@ -342,7 +357,7 @@ class IncrementalVideoStitcher:
             # 🚀 Retornamos True al final
             return (final_tensor, audio, True)
         else:
-            print(f"   -> ⏳ Ciclo intermedio. Soltando 1 frame dummy...")
+            print(f"   -> ⏳ Ciclo intermedio ({IncrementalVideoStitcher._autonomous_frame_count}/{expected_total_frames}). Soltando 1 frame dummy...")
             dummy_frame = images[-1:].clone()
 
             # 🚀 Retornamos False al final