EMA-VFI Segment Interpolate

The EMA-VFI segment node for interpolating long clips on a modest card

By ethanfel·Created 7 months ago·Updated about a month ago· 17

images
model
settings

images
model
elapsed_seconds

◄multiplier2►

◄clear_cache_after_n_frames10►

◄keep_devicetrue►

◄all_on_gpufalse►

◄batch_size1►

◄chunk_size0►

◄source_fps0.00►

◄target_fps0.00►

◄segment_index0►

◄segment_size500►

The plain EMA-VFI Interpolate node is already light on VRAM, but that doesn't save you when the input is a 20-minute video - the assembled output has to live somewhere, and that somewhere is your system RAM until it doesn't. EMA-VFI Segment Interpolate is the same interpolation split into chunks so peak memory stays bounded. If you're on a mid-range card doing long-form interpolation, this is the node that keeps the run alive.

It's structurally identical to the other pairwise Segment nodes in the pack. Same EMA-VFI internals, same multiplier, batch_size, chunk_size, keep_device, all_on_gpu, clear_cache_after_n_frames, and source_fps/target_fps - plus the two segment controls:

segment_index - which 0-based slice of the input this node processes.
segment_size (default 500) - input frames per segment, with a one-frame overlap between adjacent segments for seamless stitching.

The pattern. Drop one node per segment (index 0, 1, 2…), each ending in a Video Combine save. Chain them by feeding the previous node's model output into the next node's model input - that pass-through forces them to run one at a time, and each segment's frames get written to disk and freed from RAM before the next starts. It's manual, but it converts "interpolate a whole movie" into "interpolate several 500-frame chunks," which is the only way some machines handle it. Finish with Tween Concat Videos to merge the segment_*.mp4 files back together.

Install. Search "Tween" in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Ethanfel/ComfyUI-Tween
pip install -r requirements.txt

Restart, let Load EMA-VFI Model pull the checkpoint, and have VideoHelperSuite on hand for the video load/save nodes.

Common issues. Keep the loader's frame rate, the node's source_fps/target_fps, and Video Combine's frame_rate in agreement, or you'll get timing drift between segments that shows up as a hiccup at every stitch. If you're downsampling and the node says a segment contains no target timestamps, segment_size is too small to span one output interval - raise it. And never let two segment nodes share the same source without chaining the model output; they'll run in parallel and quietly recreate the RAM problem you were trying to escape.

Categoryvideo/EMA-VFI

Inputs (13)

Name	Type	Default	Description
images	IMAGE		Input image batch. Output frame count: 2x=(2N-1), 4x=(4N-3), 8x=(8N-7).
model	EMA_VFI_MODEL		EMA-VFI model from the Load EMA-VFI Model node.
multiplier	COMBO	2	Frame rate multiplier. 2x=one interpolation pass, 4x=two recursive passes, 8x=three. Higher = more frames but longer processing.
clear_cache_after_n_frames	INT	101–100	Clear CUDA cache every N frame pairs to prevent VRAM buildup. Lower = less VRAM but slower.
keep_device	BOOLEAN	true	Keep model on GPU between frame pairs. Faster but uses more VRAM constantly. Disable to free VRAM between pairs (slower due to CPU-GPU transfers).
all_on_gpu	BOOLEAN	false	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	11–64	Number of frame pairs to process simultaneously. Higher = faster but uses more VRAM. Start with 1, increase until VRAM is full.
chunk_size	INT	00–10000	Process input frames in chunks of this size (0=disabled). Bounds VRAM usage during processing but the full output is still assembled in RAM. To bound RAM, use the Segment Interpolate node instead.
source_fps	FLOAT	0.000–1000	Input frame rate. Required when target_fps > 0.
target_fps	FLOAT	0.000–1000	Target output FPS. When > 0, overrides multiplier and auto-computes a power-of-2 oversample up to 8x, then selects frames. 0 = use multiplier.
segment_index	INT	00–10000	Which segment to process (0-based). Bounds RAM by only producing this segment's output frames, unlike chunk_size which bounds VRAM but still assembles the full output in RAM. Chain the model output to the next Segment Interpolate to force sequential execution.
segment_size	INT	5002–10000	Number of input frames per segment. Adjacent segments overlap by 1 frame for seamless stitching. Smaller = less peak RAM per segment. Save each segment's output to disk before the next runs.
settingsopt	VFI_SETTINGS		Auto-tuned settings from VFI Optimizer. Overrides batch_size, chunk_size, keep_device, all_on_gpu, clear_cache_after_n_frames.

Outputs (3)

Name	Type	Description
images	IMAGE	—
model	EMA_VFI_MODEL	—
elapsed_seconds	FLOAT	—