并发提交｜Seedance 图生视频 (Image to Video)

Image-to-video is where most people actually live in video generation - you've already got a still you like, and you want the motion. This node is the concurrent-submit wrapper around Seedance 2.0 i2v, and it's the one to reach for when you have a pile of stills and want them all animated in one run rather than one slow clip at a time.

The concurrency model is the same across the pack: the node validates, pushes the job to a background video thread pool, and hands you a SEEDANCE_VIDEO_FUTURE. You wire those handles into SeedanceConcurrent_Video_Await, which waits on up to ten of them in parallel and returns the finished VIDEO values in slot order. Got five keyframes from a shoot you want as five clips? Five of these, one receiver.

The inputs that matter

• first_image - required. This is the frame that becomes images[0].
• last_image - optional, and the genuinely nice part of this node: connect one and you get a start/end-frame animation instead of an open-ended one. Great for loop-able motion.
• model - six choices: seedance-2.0-{standard,fast,mini}-i2v and the -global- versions. Same quality-tier story as the t2v node.
• prompt - optional, up to 20480 chars. You can drive an i2v mostly off the frame, but a sentence of motion direction helps a lot.
• seconds - 4 to 15, or -1 for model-decided.
• resolution - 480p through 4k plus native1080p/native4k; native tiers are Standard-only.
• ratio - adaptive follows your image; that's the sensible default for i2v.
• generate_audio - voice-over/sound effects toggle.

Plus api_config, skip_error (placeholder video on failure), and seed - -1 random; fixed makes unchanged runs reuse cache instead of re-billing.

How it runs

Standard wrapper machinery: the first frame (and optional last frame) gets uploaded, submitted, polled, downloaded, and validated with the MP4-header check. Failures isolate per slot in a batch - raise names the dead slot and cancels pending futures; placeholder swaps in a placeholder and keeps the rest.

Installing

# ComfyUI Manager: search "ComfyUI Seedance" → install → restart.
# or:
comfy node install seedance
# or:
cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/ComfyUI_Seedance.git
cd ../.. && python -m pip install -r custom_nodes/ComfyUI_Seedance/requirements.txt

Just requests; no weights. API key via Seedance API Config, SEEDANCE_API_KEY, or config/.env.

Where people trip

• No first_image - bounced at validation, before the API. This node has zero image-free modes.
• Native resolutions on fast/mini - Standard-tier only, same rule as the t2v node.
• Unchanged re-runs still costing money - you left the seed on randomize. Flip the submit nodes to fixed and cache reuse kicks in.

And the standing reminder: metered per clip, your frame leaves the machine, and result links are temporary. Save the keepers.