HappyHorse 1.1 is Alibaba's video model that renders picture, sound, and lip-synced dialogue in one pass. Write a prompt or upload up to 9 reference images and get a 5 to 15 second clip at 720p or 1080p. HappyHorse 1.0 is one click away in the same picker.
Select video duration
Select resolution
Select aspect ratio
CAPABILITIES
Feature 01
Picture and Sound From One Prompt
HappyHorse generates the video and its audio together. Alibaba says the model outputs lip-synced dialogue, ambient soundscapes, and expressive vocal performances in the same pass, so a clip arrives with voices and room tone instead of silence. Put spoken lines in quotes in your prompt and it times the mouth movement to them. Version 1.1 also tightened the audio-visual sync over 1.0.
Feature 02
One Photo or Nine References
Upload one photo and HappyHorse animates it, or upload up to 9 reference images and the model combines the people, products, and places from them into one scene. Alibaba's main goal for version 1.1 was reference consistency: the upgrade makes a product, a face, or an outfit from your reference hold up much better in the output. On the Artificial Analysis image-to-video arena, HappyHorse 1.1 ranks seventh at 1104.
Feature 03
Action That Holds Together
Alibaba rebuilt motion modeling and temporal consistency for HappyHorse 1.1, and the result shows in fast action: a fight, a dance, or a sprint keeps its limbs and props intact from frame to frame. The model also keeps physical cause and effect believable, which is why Alibaba pitches it for game trailers and action-heavy ads.
Feature 04
Prompt in the Language You Think In
Alibaba's API documentation says HappyHorse accepts prompts in any language, up to 5,000 characters (2,500 for Chinese). You can write the scene in Spanish, Japanese, or Chinese and the model follows it. Third-party sites claim it lip-syncs seven languages; Alibaba has not published a language list for speech, so test your language with a short clip before a long render.
Feature 05
What the Speed Claim Means
The 8-step figure comes from community write-ups that say HappyHorse uses distillation to cut denoising to eight steps. Alibaba has not confirmed it, and its own API guide says a render takes several minutes. In practice, plan on a short wait per clip, and draft at 720p with 5 seconds, which is the fastest and cheapest setting on Ricebowl.
Feature 06
Clear Speech, Matched Mouths
WER, or word error rate, measures how many spoken words come out wrong. Alibaba has not published a WER number for HappyHorse, so treat any specific figure you see elsewhere as unverified. What Alibaba does state is that the model delivers lip-synced dialogue, and its launch demo has two actors trading full sentences in a café. For clean speech, keep each line short, name who says it, and give one speaker per shot.
Feature 07
Several Shots in One 15-Second Render
HappyHorse can cut one prompt into several shots inside a single clip of up to 15 seconds at 1080p. Alibaba highlights multi-shot consistency: characters stay in position across frequent cuts, which suits short dramas built on tension and close-ups. Describe the shots in order, such as a wide on the café, a close-up on her, then him, and the model edits the scene for you.
Feature 08
Direct the Lens in Words
There are no camera sliders in HappyHorse; you direct the lens in the prompt. The model is tuned for cinematic framing with wide apertures, shallow depth of field, and film-like texture, so words like 'slow dolly-in', '35mm look', and 'rack focus to the cup' change the shot. Choose from nine aspect ratios on the official API, and landscape or portrait on Ricebowl.
Feature 09
What Is Actually Open Today
HappyHorse is not open weights today. When the model first appeared, an unofficial site described it as fully open source with GitHub and Hugging Face links marked 'coming soon'; WaveSpeed checked on April 8, 2026 and found no repository, weights, or license. Alibaba now offers it through its Model Studio API, happyhorse.com, and the Qwen App. If you need weights you can download, MiniMax H3 is open weights and runs in the same Ricebowl picker.
PROMPT EXAMPLES
Prompt
The kettle in [Image 1] sits on a marble counter at sunrise. Slow push-in as steam curls from the spout, soft window light, the rising whistle and a gentle click as the switch flips off, 16:9, 5 seconds.
Reference
Result
Prompt
A young woman pours water from the kettle in [Image 1] into a mug, looks at camera and says: "Ninety seconds. That's all my morning needs." Warm kitchen, handheld medium shot, then a close-up of the mug, 9:16, 10 seconds.
Reference
Result
USE CASES
Alibaba reports the model is already used in short-form dramas, e-commerce advertising, brand marketing, and CG game trailers. For dramas, multi-shot scenes with lip-synced dialogue let one render cover a full exchange between two characters.
Upload product shots as references and the model keeps the label and shape intact, with sound already in the clip.
Try this use case →A 1080p clip in portrait costs 145 credits for 5 seconds, cheap enough to test several hooks per campaign.
Try this use case →HappyHorse 1.1's improved action motion suits fights, chases, and reveals for pitch trailers.
MARKET CONTEXT
Current T2V and I2V leaderboard context
WHAT USERS SAY
01
Creators keep coming back to HappyHorse for scenes where people talk. Getting speech and lip movement in one HappyHorse render saves a dubbing pass.
02
Builders who went looking for HappyHorse weights found only 'coming soon' links. HappyHorse is an API and app product, not a local model.
03
Product and character fidelity was the main complaint about HappyHorse 1.0; Alibaba targeted it directly in HappyHorse 1.1.
04
HappyHorse debuted at the top of the arena in April; newer models from Google, Alibaba's Wan team, and MiniMax have since passed it.
VERSIONS
HappyHorse 1.1 succeeded HappyHorse 1.0 in June 2026, and on Ricebowl it is both better and cheaper. HappyHorse 1.1 scores 1147 on the Artificial Analysis text-to-video arena against 1119 for HappyHorse 1.0, and 1104 against 1082 on image-to-video. A 5-second 720p HappyHorse 1.1 clip costs 115 credits here; the same HappyHorse 1.0 clip costs 140, and at 1080p it is 145 against 240. There is no reason to pick HappyHorse 1.0 on this site unless you are matching an old project.
Version 1.1 leads 1.0 by 28 Elo points on the Artificial Analysis text-to-video arena (September 2026).
From a photo the gap is 22 points, with 1.1 in seventh place and 1.0 in tenth.
A 5-second clip costs 115 credits at 720p and 145 at 1080p on 1.1, against 140 and 240 on 1.0. The newer version is cheaper at every length.
WHAT SETS IT APART
01
HappyHorse entered the Artificial Analysis arena anonymously in early April 2026 and took first place on both text-to-video and image-to-video before Alibaba claimed it. In June, Arena.ai still ranked HappyHorse 1.0 second across its three video boards. Those were blind side-by-side votes, not a vendor benchmark. HappyHorse is no longer first, but that is why the name spread.
02
HappyHorse accepts up to 9 reference images and still renders audio in the same pass, a combination few models here offer at 115 credits.
03
HappyHorse 1.1 at 1080p costs 145 credits for 5 seconds on Ricebowl, well below Seedance 2.0 at 510 for the same length and resolution.
OUR TAKE
HappyHorse 1.1 is better and it costs fewer credits here, so the question answers itself. It leads HappyHorse 1.0 on both arena boards and is cheaper at every length: 115 against 140 credits for 5 seconds at 720p. Always pick HappyHorse 1.1.
Not on quality. Seedance 2.0 already outscores HappyHorse 1.1 on both boards (1210 against 1147 on text-to-video), and Seedance 2.5 is not ranked yet. HappyHorse wins on cost and resolution: 115 credits for 5 seconds at 720p against 315 for Seedance 2.5, plus 1080p, which Seedance 2.5 does not offer here. Pick HappyHorse for affordable 1080p with dialogue; pick Seedance 2.5 for clips longer than 15 seconds or more than 9 references.
HappyHorse is good for dialogue-driven short dramas, product ads from reference photos, and 1080p social clips on a budget. It is not the pick for 4K, clips over 15 seconds, the single most realistic frame (use MiniMax H3 or Seedance 2.0), or running a model on your own hardware.
COMPARISON
| Feature | HappyHorse 1.1 | Seedance 2.0 | Kling 3.0 | Wan 2.6 |
|---|---|---|---|---|
| Text-to-video arena Elo | 1147 | 1210 (720p) | 1095 (1080p Pro) | 1030 |
| Image-to-video arena Elo | 1104 | 1175 (720p) | 1054 (1080p Pro) | 889 |
| Clip length on Ricebowl | 5, 10 or 15 s | 5, 10 or 15 s | 5, 10 or 15 s (15 s on Kling 3.0 Turbo) | Not on Ricebowl |
| Max resolution on Ricebowl | 1080p | 1080p | 4K (Omni) | Not on Ricebowl |
| Sound | Built in | Built in | Optional, costs extra | Not on Ricebowl |
| Reference images | 9 | 9 | 2 | Not on Ricebowl |
| Cheapest clip on Ricebowl | 115 credits | 95 credits (480p) | 70 credits (Fast, no sound) | Not on Ricebowl |
| Best for | 1080p dialogue scenes on a budget | Highest quality of the four | 4K product shots | Older Alibaba baseline |
Arena scores are from the Artificial Analysis text-to-video and image-to-video leaderboards, September 2026. Credit prices are Ricebowl's current rates.
Try HappyHorse Now01
The generator at the top of this page already has HappyHorse 1.1 selected. Sign up with email or Google to start.
02
Describe the action, the camera, and any spoken lines in quotes. Add one photo to animate it, or up to 9 reference images and refer to them as [Image 1], [Image 2].
03
Pick 5, 10, or 15 seconds, 720p or 1080p, landscape or portrait, and press generate. The HappyHorse clip lands in your feed with its sound.
REVIEWS
Feedback from people who use HappyHorse on Ricebowl for real work.
Wei C.Two characters, three cuts, real dialogue, one render. HappyHorse replaced my dubbing step.
Short-drama director
Rachel P.Our product references finally stay on-model. The label reads correctly in every take.
E-commerce manager
Diego M.1080p with sound at 145 credits means I can pitch three versions instead of one.
Social ads freelancer
Yuki T.Fight choreography holds together. HappyHorse 1.1 is my go-to for action beats.
Game trailer editor
Amara O.We write prompts in French and the scenes come back right. No translation pass.
Brand content lead
One balance for every Rumpelstiltskin AI video you make. Pick a plan by how many tiptoe clips you want, nothing else changes between plans.
Save 30% billed annually
Flexible monthly subscription.
The cheapest way to start. Every model, a small monthly budget.
$12.99/mo
Renews monthly, cancel anytime.
400 credits
≈ 1–20 videos, depending on length
≈ 28–133 images, depending on the model
3.25¢ / credit
A few sneaky tiptoe clips a month for you and your friends.
$29.9/mo
Renews monthly, cancel anytime.
1,000 credits
≈ 3–50 videos, depending on length
≈ 71–333 images, depending on the model
2.99¢ / credit
2 parallel tasks
Steady output for a creator account riding the trend.
$49.9/mo
Renews monthly, cancel anytime.
2,000 credits
≈ 6–100 videos, depending on length
≈ 142–666 images, depending on the model
2.50¢ / credit
3 parallel tasks
High volume for pages and agencies posting every day.
$89.9/mo
Renews monthly, cancel anytime.
5,000 credits
≈ 15–250 videos, depending on length
≈ 357–1666 images, depending on the model
1.80¢ / credit
5 parallel tasks
Pay yearly and save more.
A few sneaky tiptoe clips a month for you and your friends.
$19.9/mo$29.9
Billed yearly at $238.8. Renews yearly, cancel anytime.
Save $120/yr vs monthly
12,000 credits
≈ 38–600 videos, depending on length
≈ 857–4000 images, depending on the model
1.99¢ / credit
2 parallel tasks
Steady output for a creator account riding the trend.
$34.9/mo$49.9
Billed yearly at $418.8. Renews yearly, cancel anytime.
Save $180/yr vs monthly
24,000 credits
≈ 76–1200 videos, depending on length
≈ 1714–8000 images, depending on the model
1.75¢ / credit
3 parallel tasks
High volume for pages and agencies posting every day.
$62.9/mo$89.9
Billed yearly at $754.8. Renews yearly, cancel anytime.
Save $324/yr vs monthly
60,000 credits
≈ 190–3000 videos, depending on length
≈ 4285–20000 images, depending on the model
1.26¢ / credit
5 parallel tasks
Top up credits valid for 12 months.
One-time top-up — perfect for occasional needs.
$49.9
One-time purchase, valid 12 months.
1,000 credits
≈ 3–50 videos, depending on length
≈ 71–333 images, depending on the model
4.99¢ / credit
Larger top-up with 20% bonus — best per-credit value for one-time buyers.
$198
One-time purchase, valid 12 months.
4,800 credits
≈ 15–240 videos, depending on length
≈ 342–1600 images, depending on the model
4.13¢ / credit
5 parallel tasks
Browse all
Browse all
Your pet takes the mic
Get caught on the big screen
Hold up the red sign and jump
Take the cash, then hit the walk
Walk out to a crowd of fans
A ghost rises right behind them
Walked out by two officers, unbothered
Lean in, arms out, point on the beat
Serious game face, then the toy car
Snap the moon, reveal the witch
Ask the phone how to get rich, hear the answer
Your cat or dog stands up and dances
Wide squat, tiptoes, one-leg freeze
Swagger down the park path, point at the lens
Cup your ears and let the stadium roar
Walk the carpet, stop for the flashes
Bedroom horns, panels shut, night street
Tiny workers wash your truck until it shines
Fake the rabona, leave the defender on the floor
Rocket thrusters fire, your car lifts off
Set up a penny-farthing chase through a palace garden with a golden carriage.
Create a two-person autumn field bounce and finger-point dance.
Create a two-person autumn field reunion clip with a spoken beat and cheering.
Your partner turns zombie, then a warm memory
A loved one turns zombie, hugs you, then a warm memory
Turn your photo into the Ozil goal: dribble, strike, turn and celebrate.
Turn photos into phonk head bop trend videos in minutes.
One photo, a camera orbit around your subject
One or two photos into the vertical sidewalk singing clip
Chandelier party video from your selfie photos
Turn any photo into the tiptoe trend
Create scream-and-swap pet dance clips
Turn your selfie into an animatronic jumpscare
Make your photo lead the 1983 zombie dance in 9:16 vertical video.
Turn one adult photo into a trending 9:16 lip-sync clip in seconds.
Turn one group photo into a crew that steps in sync in 9:16 vertical video.
Turn your photo into a 1980s retro dance video with custom themes.
Pop the bottle from one photo and freeze it black and white.
Turn your photo into a Bulin 47 AI doorway clip.
Do the Layoo dance from one photo with two friends behind you.
Turn your photo into the viral lean, swing and low bounce, friend watching.
Black liquid rebuilds your car from its badge
Put your photo on the red stage doing the 2008 electro arm dance.
Sweep the 2 and 4 cards, reveal a car
Deadpan sway clip from one photo
Costume walk clip from one photo.
Make your pet or baby scold the camera
Make viral sneaky tiptoe videos from photos
Walk out with two officers
Two photos turned into a drill rap video clip.
Turn your pet into a tiny knight in a storm
Turn photos into the hop and knock dance
Switch wardrobe under spinning fan blades
Turn portrait photos into video
Make the viral video from two photos
Turn one pet photo into the stage meme
Dusk road-trip clip from your photos
You and a friend trading bars on one mic
Turn their photo into a funny birthday prank video
One photo zooms out to Earth, Moon or galaxy
Laser eyes, car jumps, hugs and more from one photo
Turn one photo into a video with copies of you
Hug the kid you were in a short video
Turn family photos into a short video clip.
Turn your photo into a cinematic video with effects
Share an imagined hug with someone who passed away
Turn one or two photos into a dance video
Two solo photos, one orange booth, one mic
Sit still while a car flies over you
Turn two photos into a hug video
Turn one or two photos into a gentle memorial clip
Make people in old photos smile, wave or hug
One or two photos into the zombie love story video
Two photos into the viral Hotel Lobby video
Joining Ricebowl is free and gives you 8 credits with no card, but that is not enough for any video here: the cheapest HappyHorse 1.1 render is 115 credits (5 seconds at 720p), and even a 6-second MiniMax H3 clip costs 20. Choose a credit pack or plan to render HappyHorse clips.
HappyHorse is a video generation model from Alibaba's ATH unit. HappyHorse turns text, a photo, or several reference images into a clip with synchronized sound and lip-synced dialogue, up to 15 seconds at 1080p. HappyHorse 1.0 launched in April 2026 and HappyHorse 1.1 followed in June.
HappyHorse 1.1 improves consistency with reference images, motion in complex action, instruction following, visual detail, and audio-visual sync. It also scores higher than HappyHorse 1.0 on both arena boards, and on Ricebowl it is cheaper at every length.
HappyHorse puts dialogue, multi-shot editing, and 9-image references in one render at a low 1080p price. It is the combination at that price, from 115 credits, that sets HappyHorse apart.
No. Seedance 2.0 scores higher on both boards: 1210 against 1147 on text-to-video and 1175 against 1104 on image-to-video. HappyHorse 1.1 is the cheaper route to 1080p: 145 credits for 5 seconds against 510 for Seedance 2.0 at 1080p.
Yes. Alibaba describes HappyHorse output as synchronized audio-visual, including lip-synced dialogue, ambient sound, and vocal performance, generated in the same pass as the picture.
Alibaba's docs say HappyHorse prompts can be in any language. Alibaba has not published a list of languages for spoken dialogue, so test your language with a short HappyHorse clip first.
Up to 15 seconds per clip. On Ricebowl you choose 5, 10, or 15 seconds; a 15-second HappyHorse 1.1 clip costs 340 credits at 720p and 435 at 1080p.
No. There are no public HappyHorse weights, code, or license. HappyHorse is available through Alibaba's Model Studio API, happyhorse.com, the Qwen App, and services like Ricebowl.
Alibaba's API offers 480p, 720p, and 1080p. Ricebowl offers HappyHorse at 720p and 1080p, in landscape or portrait.
Ricebowl, because every alternative sits next to HappyHorse in one generator. MiniMax H3 ranks higher and starts at 20 credits, Seedance 2.0 ranks higher still on image-to-video, and Kling 3 Omni goes to 4K. Render one prompt on HappyHorse and a rival, keep the better clip, and pay from one balance.
RELATED
MORE MODELS
on Ricebowl!
HappyHorse 1.1 with sound and lip-synced dialogue, 9 reference images, and 1080p clips up to 15 seconds, next to every other top video model.
Try HappyHorse NowFREE SIGN-UP · SOUND BUILT IN · UP TO 1080P · 5 TO 15 SECONDS
Experience Cinematic AI Creation with HappyHorse 1.1
Try HappyHorse Now