Concurrent Submit | zhenzhen-luma-ray-v3.2-fal

Luma Ray 3.2 is one of those closed video models that ComfyUI people keep orbiting because the loop and keyframe control is genuinely good - a looping video of a waterfall or a rotating product shot just works. This is the Concurrent Submit twin of zhenzhen-luma-ray-v3.2-fal, so you get all of that plus the pack's shared video pool: fire a row of them, let the pool chew through ten at a time, collect them all with one node.

The inputs that matter

The full original input set is here:

• prompt - required.
• mode - text-to-video vs image-to-video.
• image / image_url + image_way - first-frame reference for image-to-video.
• duration - seconds for the clip.
• aspect_ratio - including portrait for Reels-style work.
• resolution - quality/resolution tier (this affects fal pricing, so watch it).
• loop - the Ray signature move. Loop the end back to the start for seamless cycles.
• keyframes_json / keyframe_indexes_json - advanced: pass explicit keyframe definitions if you're doing choreographed camera work.
• reference_image1–4 / reference_image_urls - up to four character/style references. This is what makes Ray good for "same subject, different shots" batches.
• exr_export / hdr - pro output options if you want HDR or an EXR export for grading.
• poll_interval / max_poll_attempts - fal polling control.
• api_key, seed, skip_error - the usual pack trio.

The concurrent flow

Output is a task (COMFLY_VIDEO_FUTURE), not a video - that's the deal with every Submit node. Stack your jobs, wire each task into ComflyConcurrent_Video_Await (10 slots), and it waits for all of them, hands back the finished videos in slot order, and appends a status JSON so you can see which slot died and why. Want a longer queue than 10? COMFLY_VIDEO_PENDING adds pending slots.

Install and gotchas

Part of T8mars/Comfyui-zhenzhen - ComfyUI Manager → search "Comfyui-zhenzhen", or cd ComfyUI/custom_nodes && git clone https://github.com/T8mars/Comfyui-zhenzhen then restart. No model downloads; standard deps. It's a paid API from the Zhenzhen store, and because this is a fal model it's pre-charged then settled - a resolution bump or a longer duration changes the final bill, so don't panic when an in-flight job shows a charge. And it's an overseas server: from China you'll want a proxy, per the README's very direct instructions.