PikaH3FlexReferenceVideoNode - ComfyUI Built-in Node Documentation

This node generates or edits a video with synchronized audio, guided by reference images, videos and audio clips. Pika H3 Flex is Pika’s post-trained MiniMax H3 model. Connect references and address them in the prompt as @Image1, @Video1 or @Audio1 in connection order. At least one reference image or video is required, and a maximum of 12 references can be connected in total.

Inputs

Parameter	Description	Data Type	Required	Range
`reference_images`	Up to 9 images, addressed in the prompt as `@Image1` to `@Image9` in connection order. At least one image or video is required.	AUTOGROW	No	0 to 9 images
`reference_videos`	Up to 3 clips, addressed in the prompt as `@Video1` to `@Video3` in connection order. Each clip must be 2 to 15 seconds and the clips must total 15 seconds or less. Billed per second.	AUTOGROW	No	0 to 3 videos
`reference_audios`	Up to 3 clips, addressed in the prompt as `@Audio1` to `@Audio3`. Each clip must be 2 to 15 seconds and the clips must total 15 seconds or less. When copied to the output, clips play in connection order, ahead of the sound of connected videos.	AUTOGROW	No	0 to 3 audio clips
`prompt`	What to generate, or the edit to make. Refer to references as `@Image1`, `@Video1`, `@Audio1`, or as `<Picture 1>`, `<Video 1>`, `<Audio 1>`. Must not be empty. Default: "" (empty string).	STRING	Yes	1 to 8000 characters
`reference_video_audio`	Copy the sound of connected videos to the output. Turn it off for new ambience or for motion transfer with new music. A talking head with no dialogue in the prompt then gets invented speech. Default: True.	BOOLEAN	Yes	true false
`resolution`	Output resolution. `"1080p"` and above render at 768p and upscale once. The same seed at 480p gives a different video. Default: `"768p"`.	COMBO	Yes	`"480p"` `"768p"` `"1080p"` `"2k"` `"4k"`
`aspect_ratio`	Output aspect ratio. `"auto"` follows a reference video, otherwise the prompt. Reference images do not affect it. Default: `"auto"`.	COMBO	Yes	`"auto"` `"21:9"` `"16:9"` `"3:2"` `"4:3"` `"5:4"` `"1:1"` `"4:5"` `"3:4"` `"2:3"` `"9:16"` `"9:21"`
`duration`	Clip length in seconds. The output length follows a frame grid: 5 seconds gives 124 frames, about 5.17 seconds. Default: 5.	INT	Yes	5 to 15 seconds
`rewrite_prompt`	Let Pika expand the prompt first. The same seed then gives a different video on each run. Default: False.	BOOLEAN	Yes	true false
`seed`	Seed for sampling. With `rewrite_prompt` off, the same seed repeats the video exactly, though a local run with the same seed gives a different result. Default: 42.	INT	Yes	0 to 2147483647

Constraints:

At least one reference image or video is required.
A maximum of 12 references can be connected in total, counting images, videos and audio clips together.
Each reference video and audio clip must be 2 to 15 seconds; the videos must total 15 seconds or less, and the audio clips 15 seconds or less.
The prompt must be between 1 and 8000 characters.
Referencing a slot that is not connected, such as @Image3 with two images supplied, raises an error.

Outputs

Output Name	Description	Data Type
`VIDEO`	The generated or edited clip with its synchronized audio.	VIDEO

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): c6d8700e3bb0c4ed7663aff55d477314af566568bd6db353fc21d6249ddd93bd

​Inputs

​Outputs

Inputs

Outputs