Concurrent Submit | zhenzhen-pixverse-v6-fal

PixVerse v6 is the video model with the most settings on this shelf, and this node exposes most of them: resolution, duration, negative prompt, style presets, an audio toggle, a multi-clip toggle, and a thinking mode. Wrapped in Zhenzhen's concurrent layer, it becomes a "queue twenty clips and come back" machine instead of a one-at-a-time wait.

The one thing to keep straight: it's a video node, so it lives in the shared video pool (10 workers) and returns a COMFLY_VIDEO_FUTURE task - not a finished video. You collect it with Concurrent Collect Videos (ComflyConcurrent_Video_Await), which accepts up to 10 tasks, waits on all of them, and emits video_1…video_10 in slot order. The pattern never changes: submit cheap and instant, collect once.

The inputs

• prompt - a sensible cinematic default is pre-filled ("A cinematic camera move with subtle motion.") so an empty run still does something.
• image / image_url - optional first frame for image-to-video; image_way (base64 default or image_url) decides how it travels.
• resolution - 360p / 540p / 720p / 1080p. Video cost scales hard with resolution.
• duration - 1 to 15 seconds. This is the budget dial for video.
• negative_prompt - pre-filled with the usual quality complaints; leave it or customize.
• style - none, anime, 3d_animation, clay, comic, cyberpunk. This is a real style-transfer preset, one of the friendlier things about PixVerse.
• generate_audio_switch - adds sound to the clip. Video with audio is PixVerse's party trick; it's off by default, so flip it if you want it.
• generate_multi_clip_switch - multiple clip segments per run.
• thinking_type - auto / enabled / disabled; lets the model reason about the prompt before generating, at a cost in time.
• seed - fal max 65535, 0 = random.
• poll_interval / max_poll_attempts - default 600 × 6s = 1 hour.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/Comfyui-zhenzhen

Or ComfyUI Manager → search Comfyui-zhenzhen → restart, then drop your token into api_key (ai.t8star.org overseas; api.seedance.nz for the domestic discount).

Gotchas

• 1080p × 15s × 10 concurrent jobs will empty a wallet. The pack's fal lanes are pre-deduct-then-settle; the real bill depends on resolution, duration, and the extra toggles. Start at 540p/5s for tests.
• Audio and multi-clip change runtime unpredictably. Both toggles can push a job well past the 1-hour default window; bump max_poll_attempts if you're enabling them on long clips.
• Thinking mode is slow. enabled adds reasoning time to every clip. Leave it auto and only force it on when you want the quality bump and can wait.