Concurrent Submit | zhenzhen-kling-video-v3-turbo-fal

This is the overseas-route version of Kling in the Zhenzhen pack: the same model family you saw in the low-price domestic node, but hosted on fal and billed differently. And because it's a Concurrent Submit node, you can run a whole batch of them in parallel instead of babysitting one render at a time. If you're doing storyboard shots or multi-angle takes, that's the difference between a coffee break and an afternoon.

What it wraps

Comfly_kling_video_v3_turbo_fal merges the four Kling V3 Turbo endpoints - standard and pro quality, each with text-to-video and image-to-video. The Submit twin keeps every one of those knobs but adds the concurrent pool on top. A quick word on billing: this is a fal-backed model, and Zhenzhen pre-charges the estimate (the README says "FAL模型预扣3.4，生成完后多退少补" - pre-deduct, settle up after). Seeing a charge while a job is still spinning is normal, not a double-bill.

The inputs that matter

Same list as the original node:

• prompt - the one required field.
• mode - text_to_video or image_to_video.
• quality - standard or pro. Pro costs more and is slower; standard is honestly fine for most motion tests.
• duration - 3 to 15 seconds.
• aspect_ratio - 16:9, 9:16, or 1:1.
• multi_prompt_json - the interesting one. A JSON array for Kling's multi-shot storyboard; when set, prompt is ignored. This is how you get multi-shot Kling without chaining nodes.
• image / image_url + image_way (base64 / image_url) - reference image for image-to-video. base64 uploads your ComfyUI tensor; image_url skips the upload if you already have a hosted image.
• poll_interval / max_poll_attempts - how hard the node polls fal. Default 6s × 600 attempts ≈ a 60-minute timeout.
• api_key - paste your Zhenzhen key directly; this route doesn't use the api_config settings node.
• seed / skip_error - same semantics as everywhere in this pack: seed is a ComfyUI cache-control value, skip_error turns failures into silent empty results.

The concurrent flow

Output is a single task (COMFLY_VIDEO_FUTURE), not a video. Wire one or several of these into ComflyConcurrent_Video_Await (10 slots) and the collector waits for all of them, returns the videos in slot order, and gives you a status JSON. Want a deeper queue? COMFLY_VIDEO_PENDING env var adds pending slots beyond the 10 workers.

Install and gotchas

Same pack as the rest: ComfyUI Manager → "Comfyui-zhenzhen", or cd ComfyUI/custom_nodes && git clone https://github.com/T8mars/Comfyui-zhenzhen then restart. No local models, standard Python deps. The usual Zhenzhen caveats apply: overseas server means a proxy if you're in China, and fal-model price depends on resolution and inputs - a 15s pro image-to-video render is not the same line item as a 3s standard text-to-video.