✂ Frame Set Split Beginning

Frame Set Split Beginning [DVB] cuts a frame set into two pieces: the first num_entries existing frames go to the beginning output, everything after goes to other. Both come out as full frame sets, so you can keep editing either side. It's the pack's scissors, and it exists because trimming the start of a clip - removing a rough first few frames, isolating an opening segment for a loop, dropping a section before you merge or append - is one of the most common editing moves in a video pipeline.

Counting "frames" the way it means it

The one thing to internalize: num_entries counts existing frames, not index span. A set that's gappy - say indices 0, 5, 10 - split with num_entries: 1 puts index 0 in beginning and 5, 10 in other, regardless of the gaps. The README says "in existing frames" and it means it: if your set has holes, the split happens at the count of entries, not at a timeline position. The node clamps num_entries to the length of the set, so asking for more than exists just gives you everything on the beginning side and an empty other.

The outputs keep their framerates and their indices - no renumbering happens, so beginning still starts wherever the original started. If you want the trimmed remainder to read as a fresh 0,1,2,3 clip, follow it with Frame Set Reindex [DVB]. And if you need a mid-clip division, that's Frame Set Splitter [DVB], which splits evenly in half rather than by a count from one end.

Why you'd reach for it

Mostly to trim: drop the first few frames that came out rough, or peel off an opening segment to use as a standalone loop while the other output keeps flowing through the rest of the graph. Because both outputs are live frame sets, it's also a natural branch point - send beginning down a fade path and other into an append, then merge the results. The pack's composition nodes (Append, Merge, Repeat) all take frame sets, so a split is never a dead end; it's a fork in the timeline.

One detail worth remembering: indices are preserved, not normalized. If you split off the first five frames of a 0-indexed set, beginning is 0..4 and other starts at 5 - which is almost always what you want for later appends. If you trim before appending another clip, the math just works because the indices line up.

The pack-wide exit rule

Both outputs are FRAME_SETs, and standard ComfyUI nodes can't consume that type. When a split branch is done and heading to an encoder or save node, pass it through Unwrap Frame Set [DVB] to get a plain IMAGE batch and framerate. This is the same rule as every node in this pack - wire the unwrap in once, near your output, and the whole graph stops confusing you.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/alt-key-project/comfyui-dream-video-batches.git
cd comfyui-dream-video-batches
pip install -r requirements.txt

Restart ComfyUI, or search "Dream Video Batches" in ComfyUI Manager. No model downloads - splitting is list slicing over the set's frames.