🐎 Happy Horse 1.1

Alibaba's 15B model. It debuted anonymously on the Artificial Analysis Video Arena in April 2026 and took #1 in both Text-to-Video and Image-to-Video on blind human preference votes. Its signature trick: joint video + synchronized audio in a single pass, with phoneme-level lip-sync across seven languages. 1.1 adds reference-to-video with up to 9 reference images. And it's fast.

Why it stands out

• #1 ranked, blind-tested: 1333 Elo (T2V) and 1392 Elo (I2V), top of public leaderboards by real human preference.
• Joint audio-video: a unified single-stream Transformer generates picture and sound together, so audio fits the scene.
• 7-language lip-sync at phoneme level with the lowest reported word error rate (~14.6%): English, Mandarin, Cantonese, Japanese, Korean, German, French. Other languages still render, but lip-sync is trained for these seven.
• Reference-to-video, up to 9 images: lock one face and wardrobe, then regenerate the same speaker across every language.
• Fast: a ~5s 1080p clip generates in tens of seconds; the 720p tier is cheaper than 1080p.

Specs at a glance

How to access

On Venice, Happy Horse 1.1 runs through the async video API: price a job with /video/quote, submit with /video/queue, then poll /video/retrieve. Reference-to-video takes up to 9 images in reference_image_urls; you name the language and write the spoken line inside the prompt. Confirm the exact model ID against the live models list. (Alibaba's release is closed-source, cloud-API only.)

curl -X POST 'https://api.venice.ai/api/v1/video/queue' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "happy-horse-1-1-reference-to-video",
    "prompt": "Medium close-up of a presenter in a bright studio. She looks into the lens and says, in clear English: \"Privacy is a human right.\" Precise lip-sync, quiet room tone.",
    "reference_image_urls": ["https://.../face1.jpg", "https://.../face2.jpg"],
    "duration": "6s",
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "audio": true
  }'

Draft on 720p, finalize keepers at 1080p. Audio on/off changes the credit cost; run Quote before Queue.

What Happy Horse is best for

Optimal prompt pattern

Same director formula. Because audio is native and prompts can run long (2,500 chars), describe the soundscape and (for talking heads) the exact dialogue line and language for the lip-sync engine.

Pro tips

• Draft on 720p (cheaper), finalize keepers at 1080p.
• Lock the look with the reference set, then change only the line and the named language to localize across all seven languages.
• For dialogue, name the language before the quote and keep lines short for the cleanest lip-sync.
• One speaker per beat keeps the mouth match clean; have a native speaker check translations before you finalize.

Content policy