Generate Veo3.1 Video · Veo 3 lite

There's a specific moment in every video workflow where you need to know "does this shot read at all" - does the character turn at the right time, does the camera move feel right, is the audio cue landing. That's a question a cheap render answers just as well as an expensive one. Veo 3 lite is the node for that moment: the cheapest of the three Generate Veo3.1 tiers, and the one I'd leave in the graph as your default until a shot has earned a quality-tier render.

Lite also keeps the thing people come to Veo for - native audio, generated in the same pass as the picture. The tradeoff is fidelity, not capability.

How it works

Same shape as its two siblings, one word different under the hood: the node submits a Veo3.1 generation request with model: "veo3_lite", polls the KIE task until it succeeds, downloads the file and returns a ComfyUI VIDEO. Connect images and they're uploaded to KIE before the request goes out. The task_id output is the handle the Extend Veo3.1 nodes need if you want to continue the clip.

Lite (like fast) supports REFERENCE_2_VIDEO, so you can hand it one to three reference images and have the model treat them as subject references rather than literal frames. That's the mode the quality tier doesn't offer, which makes lite a genuinely useful test bed for character-led shots even if the final render happens elsewhere.

Inputs

• prompt - required, and the only one you must fill. Detailed beats short: action, camera, light, and for dialogue shots, who speaks.
• image_urls - optional IMAGE socket. One image to start from (first-frame mode), two for a transition, or one to three as references in reference mode.
• generation_type - REFERENCE_2_VIDEO (default, supported here), TEXT_2_VIDEO, or FIRST_AND_LAST_FRAMES_2_VIDEO. Left unset, the provider infers the mode from whether images exist - pick explicitly and you'll stop being surprised.
• resolution - 720p default, plus 1080p and 4k. On lite, going above 720p mostly defeats the purpose of using lite.
• duration - 4, 6, 8. Reference mode only allows 8 seconds, so a 4-second reference-mode request gets rejected rather than clipped.
• aspect_ratio - 16:9, 9:16, or Auto (which centre-crops your input image toward the closer of the two).
• watermark - defaults to MyBrand. Empty the field. This is the most common embarrassing output from any of these Veo nodes.
• enable_translation - on by default; translates the prompt to English before generation.
• enable_fallback - deprecated; docs say remove it.
• timeout_seconds, callback_url - 1200s and empty.

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

Installing

cd ComfyUI/custom_nodes
git clone https://github.com/felipederosilva/ComfyUI-KIE-Nodes-Next
cd ComfyUI-KIE-Nodes-Next && pip install -r requirements.txt
# restart ComfyUI

Manager users: search "KIE.ai Nodes Next". The pack installs requests, Pillow, numpy, PyYAML and nothing else - no video model, no CUDA, no VRAM. Key setup is a single paste at Settings → KIE.ai Nodes Next → Connection; it's validated against KIE before it replaces anything, and environment variables are intentionally unsupported, so skip the .env route.

Two things that will waste your time

First, the node is not an output node. If its video output isn't wired into something downstream - KIE • Preview Video, KIE • Save Video, any video consumer - it won't execute at all. That's not a bug, and it also means a dangling node in your graph costs nothing.

Second, on older ComfyUI builds you'll get an explicit "your ComfyUI build does not expose … VideoFromFile, update ComfyUI" error instead of a video. That's the pack asking for ComfyUI's newer video interface, and updating ComfyUI is the fix.

The rest is economics. Lite is cheap, not free, and each take is a billed submission. If you're testing ten prompts, test them at 720p and four seconds as text-to-video, then move the winner up to fast - or quality - once you know the shot works. Chain task_id into an Extend node only after you've decided the clip is worth extending; extensions are separate paid jobs.