new.ip233.com｜seedance-2.0-mini 视频生成

The first thing to know about this node: it doesn't run anything on your machine. TikpanCangyuanSeedance20MiniNode is an API wrapper for ByteDance's Seedance 2.0 Mini text/image-to-video model, reached through the new.ip233.com relay instead of ByteDance's own API. You paste a prompt, it submits a job to the relay's /v1/videos endpoint, polls until the clip is done, downloads the mp4 into your ComfyUI output folder, and hands you a VIDEO output - all without a ByteDance account or a GPU big enough to care. This is exactly the "closed model with no open weights, so you call it" pattern from the KB's API-nodes essay: local is not an option for Seedance, so the node is how it gets into your workflow.

How it works

Mechanically it's an async submit-and-poll client. The node POSTs a JSON payload (prompt, duration, aspect ratio, resolution, any reference images/videos/audio as data URIs) to new.ip233.com/v1/videos, gets a task_id back, then polls that task every 查询间隔秒数 seconds until it's done or you hit 最长等待秒数. When the job resolves it downloads the mp4 and returns five outputs: 本地保存路径 (where it saved), 任务ID, 视频云端直连 (the hosted URL), 完整日志, and 视频输出 - the actual VIDEO you wire into a video preview or save node.

The inputs that actually matter

• API_密钥 - a key from new.ip233.com (not tikpan.com; this is a separate relay). The two hint fields in the required section just tell you where to get it.
• 生成指令 - your prompt. Seedance follows detail well, so describe motion, camera, and lighting.
• 视频时长 - 4–15 seconds (default 6).
• 画面比例 and 分辨率 - 6 aspect ratios, and 480p or 720p only. No 1080p here, that's the full model's job.

Optional but worth knowing: 首帧图/尾帧图 (first/last frame), up to 4 reference images (参考图1–4), up to 3 video URLs and 1 audio URL, and 生成原生音频 (native audio, on by default - Seedance generates sound, which most open video models still can't do). The 校验HTTPS证书 toggle defaults to false on this relay, which is unusual and a bit of a yellow flag; flip it to true if your network tolerates it.

Installing it

This ships in the ComfyUI-Tikpan-Pro pack. In ComfyUI Manager search "Tikpan" and Install, then restart - or:

cd ComfyUI/custom_nodes
git clone https://github.com/htrert/ComfyUI-Tikpan-Pro

Then restart ComfyUI. No models to download, no heavy Python deps - the whole pack is HTTP clients plus image/audio plumbing. The heavy local stuff (PSD layering, segmentation models) lives elsewhere in the pack and isn't required for this node.

Where people get burned

• Wrong key. The new.ip233.com nodes need a key from new.ip233.com, not tikpan.com. Get it from the 获取密钥请访问 hint.
• It's metered. Every run spends relay credits - Mini is the cheap tier, but a 15-second 720p clip with audio still costs real money per shot. Budget before you batch.
• Patience. Video jobs are async; a 600-second default timeout means long waits, not a hang. Check the log output for poll progress if it feels stuck.
• Skip errors (跳过错误) is there for batch runs - leave it off while you debug so failures actually surface.

It's the right tool when you want ByteDance's video quality inside a ComfyUI graph and no local path exists. Just remember the whole pipeline - prompt and references included - is going to a third-party server.