PixVerse V6 Image-to-Video

What it is

PixVerse V6 image-to-video: you bring a still, it brings motion. Standard hosted I2V, except for one thing - this endpoint also takes a template_id, which turns it into a one-click video effect: the dance, the hug, the explosion, the camera gag. Pick a template, supply however many images that template needs, done.

That's the honest pitch. Prompt-driven I2V through this node is fine - good motion, 1–15 second durations, audio switch - but if you wanted plain I2V you'd be weighing it against Wan 2.2 locally, which is free after setup. The templates are what PixVerse has that your GPU doesn't, and they're why people reach for this node specifically.

How it works

The node is generated from KIE's request schema for the endpoint. image_urls is a native IMAGE socket, and because the field is plural it takes a ComfyUI batch - connect four images and all four are uploaded to KIE's storage and sent as a list. That matters for templates: the endpoint documents that the selected template determines how many images it wants, and the batch is how you supply them.

The job runs as a task, the node polls it up to timeout_seconds (1200 seconds by default), then the finished clip lands back as a ComfyUI VIDEO.

Inputs and outputs

Required:

• prompt (STRING) - 3 to 5000 characters, can't be empty. With a template, describe the intent; without one, describe the motion.
• image_urls (IMAGE) - the source still, or a batch when the template wants several.
• duration (INT, 1–15, default 5) - and here's the trap, straight from the tooltip: "Required when template_id is not passed; if template_id is passed, the video duration is fixed by the selected template, please do not pass this parameter at the same time."
• quality - 720p default, plus 360p, 540p, 1080p.

Optional:

• template_id - a long enum of PixVerse effect templates (the schema lists dozens). Once chosen, duration is decided by the template and the number of images it needs is decided by effect_type. Preview effects at the link in the tooltip before guessing.
• generate_audio_switch - off by default; syncs audio to the video content.
• generate_multi_clip_switch - off by default; multi-clip output.
• seed - 0 to 2147483647, default 123456. Same parameters plus same seed improves reproducibility.
• timeout_seconds, callback_url.

Outputs: video, url, all_urls_json, task_id, raw_json, credits_consumed, credits_left.

Install

Manager: search KIE-Nodes-Next by felipederosilva and install, then restart ComfyUI. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/felipederosilva/ComfyUI-KIE-Nodes-Next

Then Settings → KIE.ai Nodes Next → Connection → Paste / replace KIE.ai API key. Validated before saving, shown masked, with a ✅ CONNECTED status and the current credit balance. KIE_API_KEY in the environment takes priority if you're on a headless rig. Dependencies: requests, Pillow, numpy, PyYAML.

Where people get burned

Template plus duration is an invalid combination. The API rejects setting both. If you're keeping duration in the graph for the no-template path, remember to leave it out of the job when you switch to a template.

Templates have their own image count. Feed a two-image template one image, or a one-image template a batch, and you'll get an error rather than a creative interpretation. Check the template's effect_type first.

The template enum is long and opaque. Fifty-ish entries of numeric IDs; the tooltip points you at the preview page. Do that rather than bisecting by trial, because every attempt bills you.

Batch semantics are all-or-nothing. image_urls sends everything you connect. If you connect a twelve-image batch, that's twelve uploads and a confused template.

Motion is not direction. Prompt text guides the clip, but templates dominate. If you need the shot you imagined rather than the effect you selected, drop the template and prompt hard.

Video pricing, always. Duration × resolution × attempts. credits_consumed prints on the node after every successful generation; use it.