Seedance文生视频

The text-to-video sibling of the pack's Seedance image-to-video node. Same model family, same pricing reality, one big difference: you don't need a first frame. Type a prompt, pick a resolution and duration, and Seedance builds the whole clip from nothing. It's the node you reach for when you want a product b-roll, an establishing shot, or a style-move test clip and don't have (or don't want to commit to) a still to animate.

How it works

Structurally it's the simplest video node in the pack. Your prompt gets POSTed to the gateway under the model id Dreaminat2vNode (note the slightly different string - the backend is picky, but you never type it), along with resolution, aspect ratio, duration, the camera-fixed flag and a seed. The server renders the clip at 24fps and returns it as a VIDEO tensor. Everything that made image-to-video pleasant carries over: strong instruction following, professional camera moves, and smooth style compatibility. What you lose is the anchor - there's no still to ground the composition, so the first frame is entirely the model's invention.

The inputs that matter

• prompt (STRING) - the whole job. Describe the scene, the subject, and - unless you want a locked-off shot - the camera move. Seedance reads natural language well, Chinese included.
• resolution (480P / 720P / 1080p) - and the cost knob: 1080p is dramatically pricier per second than 480P.
• Size (1:1, 3:4, 4:3, 9:16, 16:9, 21:9) - aspect ratio. 9:16 for vertical product reels, 21:9 for that cinematic pan.
• duration (INT, 3–12, default 10) - seconds.
• camerafixed (BOOLEAN) - on = camera holds still and ignores movement words in the prompt; off = obey the dolly/zoom/pan directions.
• seed (INT, default 0) - reproducibility.

One output, video (VIDEO).

Installing it

Same pack as everything here - comfyui-MJAPI-party. ComfyUI Manager, search "mojieapi_party", install, restart:

cd ComfyUI/custom_nodes
git clone https://github.com/MoJIeAIGC/comfyui-MJAPI-party.git

Then register at mojieaigc.com and set your key in config.ini (leave BASE_URL = https://www.mojieaigc.com/v1/completions untouched):

[API]
KEY = your_api_key
BASE_URL = https://www.mojieaigc.com/v1/completions

Restart to pick up the key. Nothing to download.

Common issues

• The first frame is a gamble. Without an input image, composition, style and text rendering all come straight from the prompt. Iterate on cheap 480P runs first; only pay for the 1080p take once you've seen the framing you like.
• Camera drift. If your "slow push-in" came back as a handheld wobble, either the prompt was too vague or camerafixed interfered. Explicitly say "static camera" or flip the flag and describe only subject motion.
• The usual pack failure modes: error-image-with-text means key/balance/timeout, and every render is metered - video is where a session burns through credit fastest, so keep an eye on the quota the pack's settings panel shows you.