Happy Horse 1.1 drops 1080p to $0.18/second from 1.0's $0.28, adds four aspect ratios (21:9, 9:21, 5:4, 4:5), and carries over native audio, multilingual lip-sync, 3-15 second clips, and 2,500-character prompts. Both versions run on fal with pay-per-second billing.
In this guide, I'll compare Happy Horse 1.1 and Happy Horse 1.0 on fal, covering what changed from 1.0 to 1.1, and what you can do now with Happy Horse, including a lower 1080p rate and four more aspect ratios.
Research note: To conduct this analysis, I'll compare Happy Horse 1.0 to 1.1 on identical prompts.
TL;DR
Happy Horse 1.1 is the current version of Alibaba's video generation model on fal, and the update is mostly about cost and framing.
At 1080p it now runs $0.18 per second, down from 1.0's $0.28, while 720p holds at $0.14.
The aspect ratio list jumps from five to nine, adding the ultrawide 21:9, the tall 9:21, and the 5:4 and 4:5 near-square crops.
What made 1.0 work for text-to-video carries straight over: native audio, multilingual lip-sync, 3 to 15 second clips, and 2,500-character prompts.
Happy Horse 1.0 is still on fal, and I'm about to test how it compares against Happy Horse 1.1 and see how much the model has improved over the last few months.
How does Happy Horse 1.1 compare to Happy Horse 1.0?
The table below maps both versions across the dimensions that matter for a video generation build:
| Happy Horse 1.1 | Happy Horse 1.0 | |
|---|---|---|
| Price, 720p | $0.14/second | $0.14/second |
| Price, 1080p | $0.18/second | $0.28/second |
| Resolutions | 720p, 1080p | 720p, 1080p |
| Duration | 3 to 15 seconds | 3 to 15 seconds |
| Aspect ratios | 9 (adds 21:9, 9:21, 5:4, 4:5) | 5 (16:9, 9:16, 1:1, 4:3, 3:4) |
| Native audio | Yes | Yes |
| Multilingual lip-sync | Yes | Yes |
| Endpoints | Text-to-video, image-to-video, reference-to-video | Text-to-video, image-to-video, reference-to-video, edit video |
| Prompt length | Up to 2,500 characters | Up to 2,500 characters |
| Commercial use | Yes | Yes |
Where can you access Happy Horse 1.1 and Happy Horse 1.0?
Both Happy Horse 1.1 and Happy Horse 1.0 run on fal, billed per second of output, with no subscription and no minimum commitment, in a playground and through the API.
You integrate once with the @fal-ai/client SDK, and the same call shape carries across every video endpoint on fal and the 1,000+ other AI models on the platform.
Authentication, the queue, error handling, and billing behave the same way whether you call Happy Horse or anything else.
All you need is a few lines to get started:
import { fal } from "@fal-ai/client";
const result = await fal.subscribe(
"alibaba/happy-horse/v1.1/text-to-video",
{
input: {
prompt:
"A little girl walking on the road at sunset, cinematic lighting, smooth camera movement.",
},
logs: true,
onQueueUpdate: (update) => {
if (update.status === "IN_PROGRESS") {
update.logs.map((log) => log.message).forEach(console.log);
}
},
}
);
console.log(result.data);
console.log(result.requestId);
Happy Horse 1.1 vs. Happy Horse 1.0 Side-by-Side: Video Generation Tests
These five identical prompts run through both versions on fal across fast motion, physical contact, vehicles, and a product spot (yes, I decided to test them properly!).
All five run at 16:9 except the ad, which is vertical 9:16, so the only variable between the two columns is the model version:
Test 1: Running, a night foot chase (1080p, 16:9)
Prompt:
Shot 1 (wide, 0-3s): A runner in a grey hooded top bursts from a narrow doorway into a rain-slicked alley at night, neon signs bleeding red and blue across the wet ground, breath fogging in the cold, wide establishing frame.
Shot 2 (tracking mid shot, 3-6s): The camera runs alongside at the runner's pace, puddles bursting under each stride, the hood snapping in the wind, faint motion blur and shallow depth of field.
Shot 3 (low angle, 6-8s): A low shot near the ground catches the feet driving through a shallow puddle, water spraying toward the lens, neon reflections streaking past.
Audio: pounding footsteps on wet concrete, ragged breathing, a low city hum. Anamorphic, high contrast, cinematic.
Generated using Happy Horse 1.1 on fal, an AI model from Alibaba.
Generated using Happy Horse 1.0 on fal, an AI model from Alibaba.
Notes: Both models did a really good job in cinematic quality and audio-wise, although I wasn't too happy with how Happy Horse 1.1's runner started running backwards in the end.
Test 2: Fighting, choreographed combat (1080p, 16:9)
Prompt:
Shot 1 (wide, 0-3s): Two martial artists face off in a sunlit dojo, dust hanging in shafts of light from tall windows, polished wood floor, both settling into low stances, wide symmetrical frame.
Shot 2 (mid shot, slow orbit, 3-6s): One throws a spinning kick and the other ducks beneath it, the camera arcing around the pair as the movement eases into slow motion at the miss.
Shot 3 (close, 6-8s): The second fighter answers with a low sweep and both reset into a ready stance, a tight shot on planted feet and steady breath, nothing landing hard.
Audio: the snap of the gi, bare feet squeaking on wood, sharp exhales on each move. Choreographed and clean, no blood, cinematic action lighting.
Generated using Happy Horse 1.1 on fal, an AI model from Alibaba.
Generated using Happy Horse 1.0 on fal, an AI model from Alibaba.
Notes: I like how Happy Horse 1.1 played out the pause and slow motion in the middle of the clip, especially the camerawork.
falMODEL APIs
The fastest, cheapest and most reliable way to run genAI models. 1 API, 100s of models
Test 3: Kissing, an intimate two-shot (1080p, 16:9)
Prompt:
Shot 1 (wide, 0-3s): A couple in their thirties reunite on a train platform at dusk, warm tungsten light, steam drifting across a wide frame as a train sits behind them.
Shot 2 (mid close up, 3-6s): They step together and embrace, her hand rising to rest against his jaw, soft golden backlight, shallow depth of field.
Shot 3 (slow push-in, 6-8s): The camera eases in on a tender kiss as the train pulls away behind them, both faces held in soft focus, fine film grain.
Audio: distant train brakes, a muffled platform announcement, a quiet ambient murmur. Romantic, cinematic.
Generated using Happy Horse 1.1 on fal, an AI model from Alibaba.
Generated using Happy Horse 1.0 on fal, an AI model from Alibaba.
Notes: Both models did a really good job of handling that kissing scene, and I don't really have any criticism here. I'd say that I liked Happy Horse 1.1's execution of the close-up of the couple a bit better.
Test 4: Driving, a coastal run (1080p, 16:9)
Prompt:
Shot 1 (wide aerial, 0-3s): A vintage convertible carves along a coastal highway at golden hour, ocean glinting on one side and cliffs rising on the other, a wide aerial establishing the winding road.
Shot 2 (side tracking, 3-6s): The camera keeps pace alongside the car, sun flaring off the chrome, the driver's hair whipping in the wind as they check the road ahead.
Shot 3 (crane up, 6-8s): A low shot by the front wheel cranes up and back to reveal the road curving into the distance, warm lens flare crossing the frame.
Audio: engine hum, wind rush, tires on warm asphalt, faint gulls. Warm cinematic grade, film look.
Generated using Happy Horse 1.1 on fal, an AI model from Alibaba.
Generated using Happy Horse 1.0 on fal, an AI model from Alibaba.
Notes: Absolute cinema. Crème de la crème execution from both AI video generation models right here, and there's nothing to comment on from my end other than how much I love it.
Test 5: Advertising, an influencer skincare spot (1080p, 9:16)
Prompt:
Shot 1 (vertical mid shot, 0-2.5s): A beauty influencer sits at a marble vanity in a bright, minimal bathroom, soft morning light, lifting a small frosted-glass serum bottle toward the camera in a vertical 9:16 frame.
Shot 2 (close up, 2.5-6s): A closer frame on her face as she says warmly, "Morning, everyone. Quick one today, I want to show you the serum I've added to my routine," precise lip-sync.
Shot 3 (insert, 6-8s): A tight insert as she unscrews the dropper and smooths a little onto her cheek, clean focus on the product label.
Audio: her clear, friendly voice, soft room tone, no music. Natural handheld feel.
Generated using Happy Horse 1.1 on fal, an AI model from Alibaba.
Generated using Happy Horse 1.0 on fal, an AI model from Alibaba.
Notes: I wouldn't worry too much about the frosted-glass serum bottle having text that is unrecognizable, you'll be able to use a reference-to-video endpoint with both Happy Horse models to provide them with an image of your product.
How does Happy Horse 1.1 compare against Happy Horse 1.0 on pricing?
Each version bills by the second of output, and 720p costs the same on either: $0.14 per second.
The difference shows up at 1080p, where 1.1 charges $0.18 per second and 1.0 charges $0.28.
On a 10-second 1080p clip, that's $1.80 against $2.80.
Run a hundred of those in a month, and you're looking at $180 on 1.1 versus $280 on 1.0.
At 720p there's no gap at all: a 10-second clip is $1.40 on either version.
💡 If your text-to-video is mostly 1080p, moving to 1.1 cuts the per-second cost with nothing in the request body to change.
Recently Added
Get started with Happy Horse 1.0 and Happy Horse 1.1 on fal
Each Happy Horse version runs on fal with playground access and a single API, billed per second of output with no GPU setup on your side.
Happy Horse 1.1 is at alibaba/happy-horse/v1.1/text-to-video, and Happy Horse 1.0 is at alibaba/happy-horse/text-to-video.
You can test either one in the playground, then integrate through the API when you're ready.
Happy Horse 1.1 vs. Happy Horse 1.0 FAQs
What's the main difference between Happy Horse 1.1 and Happy Horse 1.0?
For text-to-video, 1.1 costs less at 1080p and covers a wider set of frames.
It runs $0.18 per second at 1080p against 1.0's $0.28, and it supports nine aspect ratios to 1.0's five.
Native audio, multilingual lip-sync, clip length, and prompt length work the same on both.
Is there a meaningful difference between Happy Horse 1.1 and Happy Horse 1.0 in terms of video generation quality?
I wouldn't say that there's a big jump in quality in what was otherwise a really solid video generation model.
I'd say that there were noticeable improvements in Happy Horse 1.1 over Happy Horse 1.0 in terms of character movement, but lip-sync and cinematic quality-wise, I would say both models did a really good job when I tested them.
Learn more about how to best use Happy Horse 1.0 and also how Happy Horse 1.0 compares against Seedance 2.0.
How much does Happy Horse 1.1 cost on fal?
Happy Horse 1.1 bills by the second of output, at $0.18 for 1080p and $0.14 for 720p.
A 10-second 1080p clip comes to $1.80, and the same length at 720p is $1.40.
You pay per second generated, with no GPU setup to manage.
Can Happy Horse 1.1 generate speech and lip-sync?
Yes, it produces native audio in the same pass as the video, so dialogue, ambient sound, and effects come back with the picture.
All you have to do is write the spoken lines straight into the prompt for talking-head and dialogue scenes.
Lip-sync is supported across English, French, Spanish, Turkish, Japanese, and more (matched to the language you write the dialogue in).
What aspect ratios does Happy Horse 1.1 support?
Happy Horse 1.1 supports nine: 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, 9:21, 5:4, and 4:5.
That's four more than 1.0, picking up the ultrawide 21:9, the tall 9:21, and the near-square 5:4 and 4:5.
The default is 16:9.
