HappyHorse Text to Video

HappyHorse Text to Video is a cloud-hosted Wan-family video model wearing a friendly name. It lives in ComfyUI's partner/video/Wan category, so think of it as "Wan, but you don't need a GPU to run it": you write a prompt, HappyHorse renders it through Comfy's API layer, and a VIDEO comes back - billed through your Comfy account, no checkpoint files, no VRAM. The KB's Wan panel is the useful backdrop: Alibaba froze open Wan at 2.2 and took the newer numbered versions behind an API, and this node is a commercial way to keep using the Wan lineage without the local install.

The notable bit for a text-to-video node: prompts support English and Chinese - the tooltip says so explicitly. That's a genuine advantage if you're working bilingually or your ideas sit better in Chinese, since many Western-facing APIs choke on non-English prompts. This is also the plainest of the three HappyHorse nodes: no reference images, no character juggling - just a prompt and a render.

The inputs

• model - a dropdown that carries the real controls with it (the dynamic-combo pattern every HappyHorse node uses). Pick happyhorse-1.1-t2v or happyhorse-1.0-t2v and the node exposes:
  • prompt - the description, English or Chinese.
  • resolution - 720P or 1080P (1.1 adds a few more ratios than 1.0).
  • ratio - the aspect ratios on offer (16:9, 9:16, 1:1, 4:3, 3:4, and on 1.1 also 21:9, 9:21, 5:4, 4:5).
  • duration - 3 to 15 seconds, 5 by default.
• seed - seed for generation; the standard partner-node disclaimer applies (it's an input, not a reproducibility guarantee).
• watermark - whether to add an AI-generated watermark to the result.

What comes out

A single VIDEO output.

Gotchas

• The prompt box hides inside the model dropdown. If you pick a model and "lose" your text field, that's the dynamic combo doing its job - expand the model's options.
• Cheaper tier ≠ wrong tier. HappyHorse 1.0 with 720P is fine for drafts; 1.1 at 1080P is the money shot. Match the tier to the stage of the work.
• It's a paid per-call service. Same credit-budget advice as every partner node: draft cheap, then spend on the keep.
• Wan-family DNA means prompt discipline. Wan models reward concrete, scene-level descriptions; vague prompts come back generic.

If you've been eyeing Wan for video but balked at the VRAM, this is the easy on-ramp: same model family, no install, and it'll happily take your Chinese-language prompts too.