Concurrent Submit | Zhenzhen Vidu Image2Video

Vidu is the Chinese video model that keeps getting better at controlled motion, and its image-to-video mode is where most people start: one reference frame, a prompt, and a moving shot comes back. This node is the concurrent wrapper around the pack's Comfly_vidu_img2video - same Vidu call, but it submits to the shared video pool (ten workers) and returns a task instead of blocking your graph. Want to animate the same still with three different motion prompts, or test six models on one image? Queue them all, collect in Concurrent Collect Videos, pick the winner.

Inputs that matter

• image - required. This is image-to-video; no image, no job.
• model - six choices: viduq2-pro, viduq2-turbo, viduq1, viduq1-classic, vidu2.0, vidu1.5. Q2-Pro is the default and the quality pick; Turbo trades quality for speed; the older ones are there for cost or familiarity.
• prompt - motion/action prompt. Optional, but a good one is most of the result.
• audio - boolean, off by default. Turn it on and Vidu generates audio for the clip.
• voice_language - seventeen languages, defaulting to 中文(普通话). Paired with voice_id, which has 303 choices. This is the "make my generated clip speak" feature: pick a language and a voice and the clip gets dialogue in that voice. The sheer size of that list is why this node is worth digging into - most video API nodes give you two voices, not three hundred.
• is_rec - record mode toggle.
• duration - 1 to 10 seconds, default 5.
• resolution - 540p, 720p (default), or 1080p.
• movement_amplitude - auto, small, medium, large. This is Vidu's motion-intensity dial.
• bgm, off_peak, watermark, wm_position - background music, cheaper off-peak billing, and the watermark placement if you enable it.
• api_key, seed, skip_error - standard.

Output is a single COMFLY_VIDEO_FUTURE task.

How it works

Standard concurrent machinery: the ComflyConcurrent.py wrapper validates your inputs, pushes the original Comfly_vidu_img2video call onto the bounded video executor (ten workers), and returns the future immediately. The Concurrent Collect Videos node waits on every task, keeps your slot order, and reports per-slot status; failure_mode: placeholder degrades a failed slot to a blank clip instead of killing the batch.

Setup

One install for the whole pack - ComfyUI Manager (search "Comfyui-zhenzhen"), or:

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

No weights to download. You need a Zhenzhen key from ai.t8star.org (overseas) or api.seedance.nz (domestic), and each clip is a paid call - the voice and music options add to the bill, so a 10-second 1080p with dialogue isn't the same price as a silent 5-second 540p.

Gotchas

Vidu's voice catalog is a rabbit hole - pick the voice by listening, because the IDs like female-shaonv don't tell you how they actually sound. movement_amplitude is the control people forget: too much large motion and faces melt, too little and the clip is a slideshow. Off-peak billing exists for a reason; if you're batching at night for cost, it's a real discount. And remember the pool: ten concurrent video jobs, shared across every video node in the pack.