工具|异步任务查询与下载
The Tikpan async task fetcher
- 📁_本地保存路径
- 🆔_任务ID
- 🔗_云端链接
- 📄_完整日志
- 🎬_视频输出
Most Tikpan video nodes (HappyHorse, Veo, Grok) submit a job and wait for the render themselves. The Task Fetcher is the one you call afterward - it takes any Tikpan async task ID, polls the cloud until the render is done, and downloads the result to your machine. Its description calls it "通用任务轮询节点," and "universal" is the operative word: it doesn't care which model spawned the task.
Reach for it in exactly two situations. First, a long render that timed out or got interrupted - your job is probably still running on Tikpan's side, and this node is the way to collect it without resubmitting (and without paying twice). Second, batch orchestration: submit several async jobs, let them cook in parallel, then feed a row of Task Fetchers their IDs. That's the pattern the pack's own async engine uses.
How it works
The mechanism is a straightforward poll-and-download loop. You give it a 任务ID (task ID) and it hits GET /alibailian/api/v1/tasks/{task_id} on the relay host, sleeping 查询间隔秒数 (default 10s) between checks. If it gets a 429 rate-limit response it backs off gracefully rather than hammering. When the task reports done it extracts the media URL and downloads the file, using 文件名前缀 (default Tikpan_Task) to name it, saving to ComfyUI's output directory. Everything is synchronous from the canvas's point of view - the node blocks until done or until 最长等待秒数 (default 600s) runs out.
Inputs and outputs that matter
The ones a beginner actually sets: API_密钥 (your sk- key), 任务ID (paste whatever the submit node returned), 文件名前缀, and the two timing knobs - 最长等待秒数 and 查询间隔秒数. The optional 中转站地址 defaults to https://tikpan.com; leave it alone unless you're using a custom relay.
Outputs are all strings except one: 📁本地保存路径 (where the file landed), 🆔任务ID (echoed back), 🔗云端链接 (the direct cloud URL), 📄完整日志 (the full polling log, useful for debugging), and 🎬视频输出 - an actual VIDEO type you can wire into a video preview or player node without touching the filesystem.
Installing
Standard pack install - Manager search "Tikpan", or:
cd ComfyUI/custom_nodes
git clone https://github.com/htrert/ComfyUI-Tikpan-Pro
Restart, and get a key from tikpan.com. No extra dependencies; this is a pure HTTP node.
Common issues
- "轮询超时" on long renders. The default 600s max wait dies on a 16-second 4K clip if the queue is busy. Bump 最长等待秒数 to 1800+ for anything high-res; the tooltip says the same.
- Resubmitting instead of fetching. If your submit node errored after the job went through, don't immediately re-run it - you'll likely double-charge. Grab the task ID and let the fetcher collect the already-running render. This is the pack's own documented advice for its recovery flows.
- Task ID from the wrong account/relay. The task ID is scoped to your key and relay. A stale ID from a different relay host returns nothing useful; make sure 中转站地址 matches where you submitted.
- It's a key-carrying network node by design, like every Tikpan API node. That's its job - just be deliberate about where the key lives and don't trust clones of the pack.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| 获取密钥请访问 | COMBO | 1 options: 👉 https://tikpan.com (官方授权Key获取点) | |
| API_密钥 | STRING | sk- | Tikpan 平台的 API 密钥,以 sk- 开头,从 https://tikpan.com 获取 |
| 任务ID | STRING | 异步生成节点返回的任务 ID | |
| 文件名前缀 | STRING | Tikpan_Task | 下载文件的前缀名,例如 HappyHorse_T2V、HappyHorse_I2V 等 |
| 最长等待秒数 | INT | 60030–3600 | 等待任务完成的最长秒数;长视频/高清建议加大 |
| 查询间隔秒数 | INT | 105–60 | 轮询任务状态的间隔秒数 |
| 中转站地址opt | COMBO | https://tikpan.com | Tikpan 中转站地址,一般保持默认即可 |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| 📁_本地保存路径 | STRING | — |
| 🆔_任务ID | STRING | — |
| 🔗_云端链接 | STRING | — |
| 📄_完整日志 | STRING | — |
| 🎬_视频输出 | VIDEO | — |