HappyHorse 1.1 vs Kling 3.0: Same-Prompt Test and Price

We gave HappyHorse 1.1 and Kling 3.0 the same three 5-second 720p prompts in October 2026: HappyHorse won the noodle stall and the perfume orbit, and Kling 3.0 landed the skate trick. HappyHorse rendered the text-to-video clips two to three times faster. At list price, a 5-second 720p clip with sound costs $0.70 on HappyHorse and $0.63 on Kling.

Need a last frame, multi-shot or 4K? Kling 3.0 has all three. HappyHorse 1.1 takes one first frame and stops at 1080P.

Both models are on OmniaKey: HappyHorse 1.1 and Kling 3.0.

How we tested

• Models and IDs on OmniaKey: happyhorse-1.1-t2v and happyhorse-1.1-i2v for HappyHorse 1.1, kling-3.0 for Kling 3.0 (Kling's own API calls it kling-v3).
• Parameters sent, exactly: happyhorse-1.1-t2v got {"duration": 5, "resolution": "720P", "ratio": "16:9"}, happyhorse-1.1-i2v got {"duration": 5, "resolution": "720P"}, Kling text to video got {"mode": "std", "duration": 5, "aspect_ratio": "16:9"}, and Kling image to video got {"mode": "std", "duration": 5}. Kling's std mode outputs 720P.
• Dates: the Kling 3.0 clips are from our October 25, 2026 test (Kling 3.0 vs Wan 3.0), run with the same prompts and the same first frame. The HappyHorse 1.1 clips are from October 26, 2026 and also appear in HappyHorse 1.1 vs Seedance 2.0.
• 3 prompts, 1 clip per model per prompt, 6 clips in all, and every task succeeded on the first try.
• The first frame for test 3 was made with Nano Banana 2.1, so neither model animated its own image.
• Time is from sending the job to the task showing succeeded, queue time included. The two models ran on different days.
• We watched every clip and compared frames at the same moments. We asked whether the action and camera move match the prompt, whether the sign and label text is right, and whether the first frame holds. Sound was checked by loudness, two AI audio models and a spectrogram.
• OmniaKey sells API access to both models, so we ran the tests through it. Every price below is Alibaba's or Kling's own list price, not ours.
• Sources: Alibaba Cloud Model Studio's pricing page and HappyHorse text-to-video API reference, plus Kling's API pricing page and API capability map, each checked October 26, 2026.

That is one clip per prompt per model, and video models vary from run to run.

Three tests, same prompts

Test 1: Skate ollie over a puddle (text to video, 16:9)

This test checks whether the board clears the water and where the camera sits.

Prompt:

Tracking shot at golden hour on an empty concrete plaza. A teenage skateboarder in a mustard hoodie rolls toward a shallow puddle, pops an ollie over it and lands cleanly on the far side. The rear wheel clips the edge of the puddle and throws a small arc of water. Low warm sun, long shadows, handheld camera following at hip height. Sound: wheels rumbling on concrete, the clack of the board landing, a splash.

• HappyHorse 1.1 (97.6 s): The hip-height camera frames the skater from the waist up for the first 1.5 s, so the board only comes into view near the puddle, where he pops a small ollie at about 2.7 s beside it rather than over it. He lands on dry concrete at about 3.1 s, and near the end his wheels cross a wet streak and kick up a little water.
• Kling 3.0 (323.8 s): The skater rolls up, ollies over the puddle and lands cleanly on the far side, and water sprays from the puddle as the board passes. The camera stays at hip height, so his head is out of frame for most of the clip, with a concrete wall behind him in warm, low sun.

Verdict: Kling 3.0. It did the move the prompt describes, over the puddle and down on the far side, while HappyHorse jumped next to the water. Both kept the camera at hip height, so both cut off the skater's head for part of the clip.

Test 2: Night noodle stall with a sign (text to video, 16:9)

This test checks the sign text, the push-in we asked for, and whether the cook's face stays visible.

Prompt:

Night street food stall, medium shot with a slow push-in. An elderly cook in a white apron tosses noodles in a flaming wok; steam and sparks rise under a single hanging bulb. Behind him a red neon sign reads NOODLES in capital letters. A customer on the right hands over a paper ticket. Wet pavement reflects the neon. Sound: sizzling oil, the clang of the wok, distant street chatter.

• HappyHorse 1.1 (100.7 s): The "NOODLES" sign is correct, and a white-haired cook in a white T-shirt and apron works a wok with flames and steam under a hanging bulb. A customer in a green jacket holds out a paper ticket at 2.5 s, the camera pushes in slowly, and by 4.6 s the cook holds the ticket and reads it with his face in full view. He stirs the noodles but never tosses them.
• Kling 3.0 (193.6 s): The "NOODLES" sign is also correct, with flames, steam and the bulb, and a tall mass of noodles rises out of the wok at the start. From about 2.5 s the bulb sits in front of the cook's face while the camera barely moves, and at 4.6 s a customer's hand passes him the ticket.

Verdict: HappyHorse 1.1. Both spelled the sign right, and HappyHorse gave the slow push-in, kept the cook's face in view and played out the whole ticket handover. Kling lifted the noodles and its track carries the street chatter the prompt asked for, but its bulb covers the cook's face and its camera barely moves.

Test 3: Perfume bottle orbit from a still photo (image to video, 16:9)

This test checks whether a product photo survives a camera move, with the cap and two lines of label text. Nano Banana 2.1 made the frame first, then both models animated it.

First-frame prompt:

Studio product photograph of a square glass perfume bottle with a gold cap standing on wet black slate. The front label is cream paper with the word AURELIA in black serif capitals and "eau de parfum · 50 ml" in small type underneath. Water droplets on the glass and the stone, a soft key light from the left, dark background, wide 16:9 frame with the bottle in the center.

Motion prompt:

The camera slowly orbits to the right around the perfume bottle. Water droplets slide down the glass and a soft beam of light sweeps across the label. Keep the bottle, the gold cap and the label text exactly as they are. Sound: a quiet drip of water.

• HappyHorse 1.1 (91.7 s, output 1286×716): Keeps the bottle, the gold cap and the "AURELIA" label, and orbits to the right so the bottle is seen at an angle by 4.9 s, with a soft band of light crossing the label at about 2.5 s. It adds drifting smoke nobody asked for, turns the liquid a deeper amber, softens the small "eau de parfum · 50 ml" line by the end, and its track carries the water drips.
• Kling 3.0 (73.5 s, output 1284×716): Keeps the bottle, the gold cap and both label lines sharp, adds nothing, and makes only a small turn to the right while a beam of light crosses the label in the second half. Its track is close to silent at −53.8 dB.

Verdict: HappyHorse 1.1, narrowly. It orbited further, showed the light beam and added the dripping water, while Kling barely turned and is close to silent. HappyHorse also added smoke and tinted the liquid, so if the product must stay exactly as shot, Kling was the safer pick.

Results by test

Sound, speed and price per clip

Sound

All six clips came back as MP4 files with an AAC sound track, and none of them has speech. HappyHorse's perfume clip has the dripping water the prompt asked for, and Kling's perfume track is close to silent. Kling's noodle-stall clip has the distant street chatter the prompt asked for, and neither check heard chatter in HappyHorse's.

Both skate clips carry board and impact sounds. Mean loudness per clip, where closer to 0 is louder:

Alibaba's docs don't mention sound for HappyHorse: its API pages and pricing page list no audio parameter and no audio price. All three HappyHorse clips came back with sound anyway. Kling's own API treats sound as a switch, sound, that is off by default and costs more per second.

Watermark

Alibaba's watermark setting is on by default and stamps a small "Happy Horse" mark, written as two words, in the lower-right corner. Setting "watermark": false removes it, and we confirmed that on a separate 3-second 480P clip on October 26, 2026. Every HappyHorse frame in the tests above carries the mark, and Kling's clips had none.

Speed

Time from sending the job to succeeded, queue included:

HappyHorse finished both text-to-video clips in a third to a half of Kling's time and lost image to video by 18 seconds, though the two runs happened on different days and queue time may differ.

Price

Alibaba bills per second of output video and charges nothing for input, per its Model Studio pricing page checked October 26, 2026. We quote its International (Singapore) prices, and the three 1.1 task IDs cost the same.

Kling bills per second too, and one Unit is $0.14 at list price, per Kling's API pricing page checked October 26, 2026.

Neither vendor prints a per-clip price, so here is the multiplication, at list price:

For OmniaKey's current rates, see the HappyHorse 1.1 and Kling 3.0 model pages.

Pros and cons

HappyHorse 1.1

• Pros: faster text to video, at 97.6 s and 100.7 s against 323.8 s and 193.6 s; clear camera moves; finished the ticket handover; drips on the product shot; a 480P tier at $0.35 per 5 seconds; nine aspect ratios.
• Cons: jumped beside the puddle; added smoke and changed the liquid color in the product shot; watermark on by default; one first frame only, with no last frame and no multi-shot; stops at 1080P; sound not in Alibaba's docs.

Kling 3.0

• Pros: the most exact action; left the product shot untouched; fastest image-to-video clip at 73.5 s; lower list price at 720P and 1080P with sound; first and last frame, multi-shot up to 6 storyboards, and 4K.
• Cons: slowest text-to-video clip at 323.8 s; the camera barely moved on the noodle stall; the bulb hid the cook's face; close-to-silent perfume track; sound is off by default and costs more; three aspect ratios.

Which one for which job

Run the same test through one API

This request sends the skate prompt to HappyHorse 1.1:

curl https://api.omniakey.com/v1/media/tasks \
  -H "Authorization: Bearer $OMNIAKEY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: skate-ollie-001" \
  -d '{
    "model": "happyhorse-1.1-t2v",
    "type": "video",
    "operation": "text_to_video",
    "input": {"prompt": "Tracking shot at golden hour on an empty concrete plaza. A teenage skateboarder in a mustard hoodie rolls toward a shallow puddle, pops an ollie over it and lands cleanly on the far side. The rear wheel clips the edge of the puddle and throws a small arc of water. Low warm sun, long shadows, handheld camera following at hip height. Sound: wheels rumbling on concrete, the clack of the board landing, a splash."},
    "parameters": {"duration": 5, "resolution": "720P", "ratio": "16:9", "watermark": false}
  }'

Leave watermark out to get the default mark, as in our test clips.

1. The reply is 202 with a task id.
2. Fetch https://api.omniakey.com/v1/media/tasks/{id} every few seconds until status is succeeded. A video job can take several minutes, so keep polling the same id instead of resubmitting.
3. Download the clip from assets[0].url.

For Kling 3.0, set model to kling-3.0 and parameters to {"mode": "std", "duration": 5, "aspect_ratio": "16:9"}. On OmniaKey, Kling's duration has to be exactly 5, 10 or 15 seconds.

To animate a photo, set operation to image_to_video (for HappyHorse, also set model to happyhorse-1.1-i2v) and send the frame in input.media with type set to first_frame and a url that is an https link or a data URL. Leave out ratio and aspect_ratio, because the frame sets the shape.

Open the HappyHorse 1.1 and Kling 3.0 model pages to start a run, and the video API docs for every parameter.

Common questions

Is HappyHorse 1.1 better than Kling 3.0?

It depends on the clip. HappyHorse 1.1 won the noodle stall and the perfume orbit, with clearer camera moves and a face you can see, and it rendered text to video two to three times faster. Kling 3.0 did the skate trick exactly as written and left the product shot untouched, so pick it for scripted action and for products that must not change.

Which is cheaper, HappyHorse 1.1 or Kling 3.0?

At list price, a 5-second 720P clip with sound costs $0.70 on HappyHorse 1.1 and $0.63 on Kling 3.0, and at 1080P it is $0.90 against $0.84. Kling without sound is the cheapest of all at $0.42 for 5 seconds at 720P, while HappyHorse has one price and no sound switch in its docs. Only HappyHorse has a 480P tier, at $0.35 for 5 seconds.

Do HappyHorse 1.1 and Kling 3.0 make sound?

Both do. All six clips in our test came back with an AAC track inside an MP4, and none of them has speech. Alibaba's docs don't mention sound for HappyHorse and list no setting for it, while Kling's API has a sound switch that is off by default and costs more per second.

How long can HappyHorse 1.1 and Kling 3.0 clips be?

Both vendors take 3 to 15 seconds, with 5 seconds as the default. On OmniaKey, happyhorse-1.1-t2v and happyhorse-1.1-i2v take 3 to 15 seconds, and kling-3.0 takes exactly 5, 10 or 15 seconds.

How do I turn off the HappyHorse watermark?

Add "watermark": false to the request parameters. The mark is on by default and sits in the lower-right corner, and we checked the switch on a separate 3-second 480P clip on October 26, 2026: the clip came back with no mark.

Update log

• October 27, 2026: first version; Kling 3.0 clips from October 25 and HappyHorse 1.1 clips from October 26, 2026.