FL Fal Pixverse API

Image-to-video (or text-to-video) through Pixverse v5, one of the mainstream closed-source video generators people put up against Kling and Runway. This node is a batch-capable wrapper - feed it a prompt (and optionally a starting image), and it fires up to 5 parallel generations through Fal.ai's Pixverse endpoint, then hands you the extracted frames from each.

The mechanism

Pixverse is a hosted, proprietary video model - never something you run locally - so this node packages your parameters into API calls and extracts frames from whatever video comes back. batch_size (1-5) is literally how many separate generations run in parallel per queue; each lands in its own frames_N output slot.

Inputs and outputs that matter

prompt describes the motion or scene. image (optional) supplies a starting frame for image-to-video; leave it empty for text-to-video. aspect_ratio covers 16:9, 4:3, 1:1, 3:4, and 9:16. resolution runs 360p through 1080p - a real cost/quality lever with a hosted video model, since 1080p costs more and renders slower than 360p. duration is "5" or "8" seconds. style gives six presets baked into Pixverse itself (none/anime/3d_animation/clay/comic/cyberpunk), and camera_movement offers 21 choices - pans, zooms, crane shots, rotation, and more, with "none" if you just want the subject to move without deliberate camera work. negative_prompt ships with a long, sensible default covering the usual "blurry, low quality, pixelated" boilerplate. nth_frame (1-4) controls the frame-extraction stride from the returned video - 1 gives every frame, higher values thin it out.

Outputs: frames_1 through frames_5 (one IMAGE batch per parallel generation slot - if batch_size is 3, results land in frames_1 through frames_3), video_urls (the hosted video URLs, handy if you want the actual file rather than just extracted frames), and status_msg.

Installing it

ComfyUI Manager: search Fill-Nodes, install, restart. Manually: cd ComfyUI/custom_nodes && git clone https://github.com/filliptm/ComfyUI_Fill-Nodes, then restart.

Where people get burned

Pixverse is a paid, metered service on top of your Fal.ai billing, and video generation is the most expensive category of call in this whole pack - batch_size=5 at 1080p/8s means five separate expensive jobs, not one, so start at batch_size=1 while dialing in a prompt.

Community sentiment on Pixverse itself is mixed. It's a legitimate, widely-used image-to-video service, but it's had rounds of "pricing feels nickel-and-dime" complaints and quality shakeups after platform redesigns - so if a result disappoints, it's worth comparing the same prompt on a different i2v backend before assuming your node settings are the problem.