ComfyUI Extension
ComfyUI-Tutu-360VR-Tools
Memory-conscious ComfyUI nodes used by TUTU's LTX 2.3 360 VR video workflow.
zhaotututu/ComfyUI-Tutu-360VR-Tools
Nodes—
On cloudLocal install
Stars0
Updatedabout a month ago
Readme
ComfyUI-Tutu-360VR-Tools
Memory-conscious ComfyUI nodes used by TUTU's LTX 2.3 360 VR video workflow.
Nodes
TUTU Stream Equirect Source Composite- Fits a temporally stable tone correction on the known source region.
- Re-composites the clear source video in bounded frame chunks.
- Supports wrap-aware feathering and longitude tone equalization.
TUTU In-Place Longitude Roll- Restores equirectangular longitude one frame at a time.
- Avoids duplicating a long 4K image batch in system memory.
TUTU Stream 8K 360 Video Upscale- Applies a tiled image upscaler one frame at a time.
- Uses circular horizontal padding to protect the 360 seam.
- Streams frames directly to FFmpeg instead of keeping an 8K batch in RAM.
- Uses NVIDIA H.264 NVENC when available and falls back to
libx264.
Install
cd ComfyUI/custom_nodes
git clone https://github.com/zhaotututu/ComfyUI-Tutu-360VR-Tools.git
Restart ComfyUI after installation.
Model
The 8K node accepts a standard ComfyUI UPSCALE_MODEL. The published workflow
uses 4x-UltraSharp.pth, placed in:
ComfyUI/models/upscale_models/4x-UltraSharp.pth
Runtime Notes
- FFmpeg is required for video output.
- 8K output is streamed, but the upstream 4K image batch still consumes system memory. Reduce the source frame cap when system RAM is limited.
- The in-place roll node should be the only consumer of its input image.
Attribution
The workflow also uses the public
ComfyUI-Seamless-Equirectangular
and
ComfyUI-VR-Outpaint-Tools
projects. This package contains workflow-specific memory and output nodes. It
does not claim ownership of LTX, the upstream projects, or their algorithms.