A multi-frame IMAGE output is not a VIDEO object in ComfyUI: it is a sequence of frames, while VIDEO is a distinct data type expected by nodes whose input contract says VIDEO. Use the original video when the next node requires VIDEO; split it into frames only when the next operation accepts IMAGE, and use Create Video if the processed sequence must become a video again.
Read the connection by type
ComfyUI’s Get Video Components documentation describes a VIDEO input and separate outputs: images (IMAGE), audio (AUDIO), fps (FLOAT), bit_depth (COMBO), and color_space (COMBO). These are independently usable components, not interchangeable labels for the same object. In particular, extracting frames does not make the images output a VIDEO, and it does not bundle the separate audio output into those frames.
Judge compatibility by the receiving socket’s declared type and the node’s input description, not by socket color, similar names, or the fact that the frames came from a video. A multi-frame image batch can be suitable for image operations while still being the wrong type for a node that requires VIDEO.
Route from the downstream requirement
Use this sequence before wiring a graph:
- Identify the first downstream node that will consume the media, and inspect its input type. Record whether it requires
VIDEOor acceptsIMAGE. - If it requires
VIDEOand you do not need to edit frames separately, connect the video to that input. Do not insert Get Video Components just to obtain a visually similar multi-frame output. - If the operation accepts
IMAGE, connect theimagesoutput from Get Video Components. Treat audio, frame rate, bit depth, and color space as separate outputs when the workflow needs them; do not assume they travel inside the image batch. - If the operation changes the frame sequence and the next stage again requires
VIDEO, connect the resultingIMAGEsequence to Create Video. Supply anfpsvalue and, if needed, anAUDIOinput. - Check that the created output is
VIDEObefore connecting it to a video-only node. For properties that must be preserved, verify the relevant settings and result separately rather than assuming that splitting and recombining guarantees preservation.
The Create Video documentation specifies IMAGE as its required image input and VIDEO as its output. It also provides an fps input and an optional audio input. Reassembly is therefore a type conversion path, not proof that every property of the original video has been retained. Detailed frame-rate and audio-preservation checks belong to a separate reassembly review.
Type-routing worksheet
The following is a hypothetical routing example, not a report of a tested graph. Suppose a clip is described as 48 frames at 24 fps, and the next node declares VIDEO. Passing the extracted images output to that socket is a type mismatch even though the images originated in the clip. If the next node instead accepts IMAGE, process the frames there; if a later node requires VIDEO, reassemble the resulting sequence with Create Video.
| Current data | Next input contract | Route | Check before continuing |
|---|---|---|---|
Original VIDEO | VIDEO | Keep the video connected | Confirm the receiving input is VIDEO |
Extracted IMAGE frame batch | IMAGE | Process the frame sequence | Confirm the node accepts IMAGE |
Extracted IMAGE frame batch | VIDEO | Do not connect directly; use Create Video if reassembly is intended | Confirm the new output is VIDEO |
IMAGE sequence plus a separate AUDIO output | VIDEO | Connect images to Create Video and optionally supply audio | Inspect the resulting VIDEO and any properties that matter |
When a connection fails
If ComfyUI refuses a connection, first compare the source output type and target input type. If they differ, choose a route that produces the required type rather than renaming or rearranging nodes. If the target’s type is unclear, inspect its input description or its official node documentation before choosing a conversion.
If Create Video does not produce the expected downstream input, confirm that its output socket is the one being used. Its documented codec options affect encoding behavior: none keeps the output in tensor form, while a selected codec encodes the video immediately. The documentation does not establish that either choice restores every original media property, so do not treat codec selection as a fix for a VIDEO versus IMAGE mismatch or as a guarantee of full fidelity.
Common questions
Why can’t I connect extracted frames directly to a video node? Because the images output is documented as IMAGE, and the receiving socket requires VIDEO. Shared origin does not make the types equivalent.
Does Get Video Components remove the video or its audio? It exposes frames and audio as separate outputs, along with other video properties. Use the original VIDEO path when the downstream operation needs the video object itself; route the separate components explicitly when processing them.
Will Create Video recreate the original video exactly? The documentation describes combining images into a video with a chosen frame rate and optional audio. It does not promise exact restoration of the original timing, audio, or other properties. Verify the properties required by your workflow.
Sources and next step
- GetVideoComponents - ComfyUI Built-in Node Documentation
- CreateVideo - ComfyUI Built-in Node Documentation
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the film workspace. The external software controls discussed above belong to their respective products.