01Body mechanics
Knees bend the right way
Joints, weight transfer, balance — the things generic models flub on first contact. Happy Horse was trained on dance and athletic footage, so motion reads as a body, not a puppet.
Single render · live clip
AlibabaLive in Snaptalealibaba/happyhorse-1.0
Happy Horse 1.0 people who move.
Alibaba's real-people motion model. Dance, runway, athletics, expressive performance — the kind of vertical clips that thumb-stoppers are made of. Snaptale calls Happy Horse the moment a prompt centres on a human moving.
auto-routed · refined in chat
9:16
9:16
9:16
- Real-people motion
- Native audio
- Multi-language lip-sync
- Vertical native
- Beat-aware
- Open weights
What it's actually good at, scene by scene.
02Audio + video, one pass
Sound and picture rendered together
Most video models leave you to ADR the audio after the fact. Happy Horse synthesises diegetic sound — footfall, ambience, dialogue — alongside the visual in a single pass. Snaptale's studio lets you swap the track if you want a different soundtrack on top.
Single render · live clip
03Lip-sync
Mouths that match the line
Mandarin, Cantonese, English, Japanese, Korean, German, French — Happy Horse syncs lip and jaw movement to the audio it generates, in any of those languages. Useful for shorts, explainers, and any time a face needs to actually speak.
Single render · live clip
04Image-to-video
Animate the portrait you uploaded
Pin any portrait as the first frame and Happy Horse takes it from there — the subject keeps your reference's face, posture and styling while the motion follows your brief. Bring your own protagonist.
Single render · live clip
·01
·02
·03
059:16 from frame one
Born for the format your audience scrolls
Subject framing, headroom, motion arc — all default to vertical when you ask for 9:16. No center-cropping after the fact, no head sliding out of frame on TikTok and Reels.
Three frames · same subject
03 — How it runs
Just talk. Snaptale routes.
No model picker, no provider key, no prompt engineering ritual. Describe the shot, the agent picks Happy Horse 1.0 when it fits — and you stay in the conversation while it iterates.
- in-threadrunway strut, slow dolly, golden hour, hold on the reveal turnHappy Horse 1.0 — staging your subject…01step / 03
Describe the action
Less "a woman walking", more "runway strut, slow dolly, tight crop on the shoes hitting the line". The more performance language you use, the better the result.
- renderalibaba / happyhorse-1.0queued · 9:16 · 720pgeneration · 720p · audio off02step / 03
Snaptale routes the request
The agent reads the motion verbs (dance, run, strut, jump, throw, glide) and queues Happy Horse instead of a generic video model.
- refine›snap beat: 124 BPM⌘1›swap track: Lo-fi Drift⌘2›trim head 0.4s⌘303step / 03
Tighten in the studio
Drop the clip on the timeline, snap the beat marker, replace the music. Happy Horse's pacing already gives you something to cut to.
05 — Inside Snaptale
Why run it here.
Happy Horse 1.0 is great. Snaptale makes it directable — persistent assets, character locks, the studio and the agent all sharing one project.
- Vertical · safe-zone aware
Vertical-first workflow
Snaptale's studio previews 9:16 by default, with safe-zones that mirror the platforms you'll publish to. No surprise crops at export.
- Beat · snap-to-grid
Beat-aware studio
Drop a music track on the timeline and the trim handles snap to the beats Happy Horse composed for — saves the manual nudging.
- Reel · multi-clip
Multi-clip reel pipelines
Generate 4–8 vertical clips on a single brief, line them up as a reel, swap one without re-rendering the rest.
- Brand · pinned
Brand & character continuity
Pin a person to your project once. Every Happy Horse clip in the project inherits the look, palette and identity automatically.
Quick answers.
01Is Happy Horse only for human subjects?
It excels at human motion, but it handles animals and inanimate motion too. The threshold the agent uses to route here is whether the focal motion is biological — gait, pose, performance.
02Can I lock in a specific person?
Yes. Upload or generate a portrait, pin it to the project, and reference it in your prompt. Happy Horse will treat it as the first-frame anchor.
03What duration and resolution?
Up to 5 seconds at 720p, 9:16 native. Snaptale's exporter handles the upsample to 1080p without re-rolling the model.
04Does it generate audio natively?
Yes — sound and picture render in the same pass. If you'd rather use your own track, mute the generated audio in Snaptale's studio and drop a music file on the timeline.
05How does it compare to Seedance?
Seedance shines for cinematic narrative and character continuity across multi-cut stories. Happy Horse shines for kinetic, real-people motion you want to scroll past in a feed.
06Will it deepfake a real person from a single photo?
Snaptale enforces consent: you can only pin photos you own or have rights to. The model produces likeness-accurate motion of legitimately uploaded subjects, not impersonation of arbitrary public figures.