HappyHorse T2V Config

HappyHorse is Alibaba's new video model line, and it arrived in the middle of 2026 with all the confusion that new Alibaba video announcements attract - the model's existence got tangled up with fake repos and scam mirrors before the real one even settled. HappyHorseBailianCNT2VConfig sidesteps all of that drama: it talks to the actual model through Alibaba's own Bailian/DashScope cloud, so there's nothing to download and nothing to get scammed by. The official endpoint is the only door this node knows.

It's a text-to-video config: prompt in, MEDIAHUB_VIDEO_REQUEST out, then the standard chain - MediaHubSubmitVideo → MediaHubResolveVideo → SaveVideo. Auth is a DASHSCOPE_API_KEY, the same key as the Wan Bailian nodes.

Inputs that matter

• model - one choice, happyhorse-1.0-t2v. No version picker, no fast variant; this node is deliberately simple.
• resolution - 720P or 1080P.
• ratio - the most options of any video config in the pack: nine, including the usual 16:9/9:16/1:1, the 4:3/3:4 and 4:5/5:4 pairs, plus the cinematic 21:9 and ultra-tall 9:21 for vertical shorts. HappyHorse clearly ships with social-format framing in mind.
• duration - default 5 seconds, range 3–15.
• watermark - on by default. Note the difference from the other providers, which default to off. If you don't want the HappyHorse branding/AIGC mark, flip it off explicitly.
• seed - -1 = random.

Output: request → MediaHubSubmitVideo.

Why you'd reach for it

If you want to be early on a brand-new Alibaba video model without the hassle of finding trustworthy weights, this is the clean on-ramp. The nine aspect ratios also make it the best pick in the pack if you're producing for TikTok/Shorts/Reels form factors and want exact framing choices.

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/vantang/ComfyUI-MediaHub.git
/path/to/ComfyUI/python -m pip install -r ComfyUI-MediaHub/requirements.txt

Restart ComfyUI; node under MediaHub/Provider/HappyHorse Bailian CN/Video. Requirements are requests/Pillow/numpy, and a DashScope key from Alibaba Cloud.

Common gotchas

The watermark default is the one that bites - every provider in this pack ships it off except HappyHorse, so a workflow you ported from a Seedance node will produce watermarked clips until you check this one. Also remember this is a new model: availability and quality are still settling, so treat early outputs as a first look, not a verdict. And the usual async rules apply - Submit hands you a job, ResolveVideo does the waiting, and the ten-minute poll window is normal for this kind of render.