工具｜提示词选择器·视频

The video prompt is a different beast than the image prompt - camera movement, cuts, pacing, shot length all matter, and writing one from scratch is where most people stall. The Video Selector is the Tikpan prompt library's video-scoped picker: a dropdown of video-generation cases, pre-curated. The brief shows 163 entries at the top of the list, and the visible samples tell you exactly what kind of curation this is - "街头说唱MV表演", "黑猫沙漠武术对决", "15秒原创元素战斗短片" - the viral short-video end of the spectrum, with links to the source X (Twitter) posts baked into the labels. This is a TikTok-e-commerce pack, so the video cards lean heavily on what actually performs as short content.

Like the other selectors, it's a convenience layer: no API, no key, no network. It reads the synced prompt library and serves a card. The video-scoped filter (CARD_TYPE_FILTER = "video") means the dropdown and the 过滤_仓库 options narrow to video repos - the brief shows the Seedance 2.0 prompts collection, with a seedance tag filter.

How it works

Same machinery as the general selector: card list scoped by tag, dropdown built from titles + prompt previews, filters for repo/tag/keyword, 显示详情 toggle. The notable quirk of the video cards: the labels embed source links (Case 1: [街头说唱MV表演](https://x.com/songguoxiansen/st…)), so you can click through to the original example video a case is based on. The output prompt text is the full video prompt for that case - ready to feed into a Tikpan video node (HappyHorse, Veo, Vidu3, Seedance-family, etc.).

Inputs and outputs that matter

选择提示词 is the dropdown. Outputs are the same three STRINGs as the other selectors: 提示词文本 (the video prompt - wire into a video node's prompt input), 卡片标题, and 卡片详情. 过滤_标签 here is a thin list (mostly seedance), and 过滤_仓库 shows only repos containing video cards. 显示详情 toggles detail output.

Installing

Standard pack install - Manager search "Tikpan", or git clone https://github.com/htrert/ComfyUI-Tikpan-Pro into custom_nodes, restart. And the standing prerequisite for every library node: run the Prompts Manager's 同步全部仓库 first or the dropdown will be empty.

Common issues

• Empty dropdown. Unsynced library. Run the manager once - it's always this.
• Cards are tuned for Seedance. The visible tag set and repo list make the model-flavor obvious: these cases were written against Seedance 2.0. Feeding them into Veo or HappyHorse can still work, but shot-count and pacing language is model-specific - a prompt built around Seedance's strengths won't map 1:1 (prompt-engineering.md).
• Video prompt ≠ image prompt conventions. These cards often encode timing ("15秒", "每切0.5秒", "10秒内20次硬切") that only matters if the target node respects duration framing. Match the card's stated duration to your node's video-length setting.
• {argument} templates. Same as the image cards: some expect substitution before generation. Fill placeholders, don't ship literal braces.