Files

T

Eric Rollei abffb216a2 fix: Clear Downstream node now accepts IMAGE/STRING/LATENT triggers

The wildcard '*' input type doesn't work properly in ComfyUI.
Changed to explicit optional inputs:
- trigger_image: Connect IMAGE output (most common)
- trigger_string: Connect STRING output (like filepath)
- trigger_latent: Connect LATENT output

Connect any one of these to ensure the node runs after your workflow completes.

2025-12-02 13:30:14 -08:00

1.9 KiB

Raw Permalink Blame History

Caching Hunyuan model for reuse: A:\Comfy25\ComfyUI_windows_portable\ComfyUI\models\HunyuanImage-3-INT8 ✓ INT8 model loaded - VRAM Allocated: 80.81GB, Reserved: 81.09GB Model distributed across GPU + CPU as needed ✓ INT8 model ready for generation with budget-aware nodes

LARGE IMAGE GENERATION MODE (Budget) Resolution: 1152x864 Offload Mode: disabled -> Disabled Starting Memory: RAM=2.16GB, VRAM=85.61GB / 95.26GB

Applied resolution patch to bypass bucket snapping Generating image with 40 steps Rewriting prompt using LLM API (style: en_recaption)... Calling LLM API: https://api.deepseek.com/v1/chat/completions (model: deepseek-chat) Original prompt: A detailed painting in the Italian Renaissance style, reminiscent of Hieronymous... Rewritten prompt: A detailed painting in the Italian Renaissance style, reminiscent of Hieronymous... Prompt: A detailed painting in the Italian Renaissance style, reminiscent of Hieronymous Bosch's art style. ... Applied resolution patch to bypass bucket snapping Using custom resolution: 1152x864 Guidance scale: 7.0 100%|██████████████████████████████████████████████████████████████████████████████████| 40/40 [02:35<00:00, 3.90s/it] Restored original resolution logic Converting image (mode: RGB, size: (1152, 864)) ✓ Image generated successfully - tensor shape: torch.Size([1, 864, 1152, 3]) GPU memory after generation: 95.26GB used, 0.00GB free

✓ Large image generated successfully (budget node) MEMORY: Peak VRAM 92.2GB | VRAM Start 85.6GB → End 95.3GB | Peak RAM 331.3GB

1.9 KiB Raw Permalink Blame History

Caching Hunyuan model for reuse: A:\Comfy25\ComfyUI_windows_portable\ComfyUI\models\HunyuanImage-3-INT8 ✓ INT8 model loaded - VRAM Allocated: 80.81GB, Reserved: 81.09GB Model distributed across GPU + CPU as needed ✓ INT8 model ready for generation with budget-aware nodes

LARGE IMAGE GENERATION MODE (Budget) Resolution: 1152x864 Offload Mode: disabled -> Disabled Starting Memory: RAM=2.16GB, VRAM=85.61GB / 95.26GB

✓ Large image generated successfully (budget node) MEMORY: Peak VRAM 92.2GB | VRAM Start 85.6GB → End 95.3GB | Peak RAM 331.3GB

1.9 KiB

Raw Permalink Blame History