gVideo
Alibaba · HappyHorse 1.0

The #1-ranked AI video model on Artificial Analysis — now on gVideo

HappyHorse 1.0 — Alibaba's #1-Ranked AI Video Model

Run Alibaba's HappyHorse 1.0 on gVideo alongside 11 other models — Kling 3.0, Veo 3.1, Sora 2 Pro, Luma Ray 2, Wan 2.6, Seedance 2.0, Hailuo 2.3, Kling 2.5 Turbo, and Pika 2.2. One subscription, one credit pool, no extra setup. HappyHorse leads the Artificial Analysis blind-test rankings for both text-to-video and image-to-video — and it ships with native audio synthesis baked in.

HappyHorse 1.0Audio · auto
Kling 3.0Wan 2.7Veo 3.1 FastVeo 3.1 ProVeo 3.1 LiteSeedance 2.0Sora 2 ProSora 2Kling 2.5 TurboHailuo 2.3Pika 2.2Luma Ray 2Luma Ray 2 FlashHappyHorse 1.0

What’s different

Why creators reach for HappyHorse 1.0

#1 on Artificial Analysis

HappyHorse 1.0 climbed to the top of blind-test leaderboards for both text-to-video and image-to-video generation in April 2026. Built on a 15B-parameter transformer, it ranks ahead of every other open-weights and proprietary model on the benchmark.

Native audio + lip-sync

Audio is synthesized in the same pass as video — no separate dubbing step. Speech, ambient sound, and music are all generated together, with precise lip-sync when characters speak.

Multimodal reference inputs

Combine a text prompt with 1-9 reference photos for character, product, or style consistency across the clip — or feed in an existing video and describe how to edit it. Reference and video-edit modes both run natively on gVideo.

50+ visual styles, 5 aspect ratios

Cinematic, anime, documentary, commercial, fashion, sci-fi, and 40+ more out of the box. All five aspect ratios supported (16:9, 9:16, 1:1, 4:3, 3:4) for any platform.

Sample generations

Cinematic cafe portrait
16:9
Anime · school rooftop
16:9
Macro luxury watch
1:1
Vertical · neon Tokyo
9:16
Skater kickflip · slow-mo
16:9

One prompt · 14 engines

Same prompt, 10 models

This is what aggregator access unlocks: one prompt, run in parallel against every model on gVideo — Sora 2 Pro, Kling, Veo, Luma, Seedance, Hailuo, Wan. You see the style differences side by side before committing credits to a final render.

Prompt: “A golden retriever sprinting through shallow waves at sunset, slow motion, cinematic 35mm, warm backlight

Kling 3.0
Veo 3.1
Seedance 2.0
Wan 2.7
Sora 2 Pro
Kling 2.5 Turbo
Hailuo 2.3
Luma Ray 2
Pika 2.2
HappyHorse 1.0THIS PAGE

Samples are placeholders — real gVideo generations will replace these in the next release.

Head-to-head

HappyHorse 1.0 vs the other models on gVideo

ModelShort-run creditsSpeedMax durationBest for
Pika 2.220/ 5s30–75s10sStylized · 720p default
Veo 3.1 Lite20/ 4s60–120s8sCheapest audio-native · Google Veo
Kling 2.5 Turbo25/ 5s30–60s10sFast Kling renders
Hailuo 2.330/ 6s40–90s10sChinese realism · 768p default
Wan 2.730/ 5s30–75s15sNative audio + 2-15s + multi-aspect
Luma Ray 245/ 5s40–90s9sDream Machine · base default
Kling 3.040/ 5s60–120s10sLifelike motion
Veo 3.144/ 4s90–180s8sPhotoreal + audio
Seedance 2.090/ 5s40–90s10sSpeed + adherence
HappyHorse 1.0THIS PAGE60/ 5s90–150s15s#1 benchmark · native audio · 720p default
Veo 3.1 Pro140/ 4s120–240s8sTop-tier cinematic · 4K
Sora 236/ 4s90–180s20sOpenAI cinematic · budget 720p
Sora 2 Pro164/ 4s90–180s20sOpenAI cinematic · HD default

Credits

HappyHorse 1.0 credit cost on gVideo

HappyHorse 1.0 costs 60 credits per 5-second video, or 120 credits for 10 seconds. All 14 models share a single credit pool under your gVideo subscription.

HappyHorse 1.0 Standard 720p costs 60 credits per 5-second video, with native audio baked into every render. HD 1080p (115 credits per 5s) is available when you need print-quality output, and clips run from 3 to 15 seconds. Free 100-credit signup covers a first render to test.

60
cr / 5s
120
cr / 10s

Common questions about HappyHorse 1.0

What is HappyHorse 1.0?

HappyHorse 1.0 is Alibaba's flagship AI video model, released April 2026. It topped the Artificial Analysis blind-test leaderboard for both text-to-video and image-to-video on launch — a first for an open-source model. Built on a 15B-parameter transformer with native audio synthesis.

Is HappyHorse 1.0 really #1 on Artificial Analysis?

Yes, as of April-May 2026. HappyHorse 1.0 climbed to the top of both text-to-video and image-to-video leaderboards on artificialanalysis.ai shortly after release. The model was unattributed at first ("the mystery #1 model") and later confirmed as Alibaba's project.

How does HappyHorse 1.0 compare to Sora 2 Pro?

HappyHorse 1.0 ranks above Sora 2 Pro on the current Artificial Analysis blind tests, while costing roughly a third of Sora 2 Pro per 5-second render on gVideo (60 credits vs 164). Sora still has the edge on very long-form coherence (up to 20s); HappyHorse runs 3–15s on gVideo.

Does HappyHorse 1.0 support audio?

Yes — natively. Audio (speech, ambient, music) is synthesized in the same pass as video, with precise lip-sync when characters speak. This puts it in the same audio class as Veo 3.1 and Sora 2 Pro.

What aspect ratios does HappyHorse 1.0 support on gVideo?

All five: 16:9, 9:16 vertical, 1:1 square, 4:3, and 3:4. More aspect ratios than any other model on gVideo.

Can I use HappyHorse 1.0 for commercial work?

Yes on all paid plans. The free tier is personal-use only. The base HappyHorse 1.0 model is open-source under a permissive license, but using it via gVideo wraps it under our standard commercial license for paying users.

Ready to generate with HappyHorse 1.0?

Start free — 100 credits on signup, no credit card required.