Happy Horse 1.1

Happy Horse 1.1 is Alibaba's upgraded joint audio-video generation model, released in June 2026. Like the original, it generates video and synchronized audio together in a single pass — including native lip-sync — but version 1.1 brings systematic improvements in dynamic expressiveness, subject consistency, instruction following, visual quality, and audio naturalness. On Morphed it costs 20 credits per second at 720p or 26 at 1080p, for 3-15 second clips.

Five areas: (1) motion is smoother and more expressive, especially for human performance; (2) characters and subjects stay more consistent through movement and cuts; (3) prompts are followed more literally, including camera direction; (4) overall visual fidelity is higher; and (5) audio is smarter — dialogue now adapts its delivery, pacing, and emotional tone to the scene context, and you can describe background and environmental sounds directly in the prompt. Formats also expanded from 5 to 9 aspect ratios.

Happy Horse generates audio and video jointly rather than layering sound onto silent footage — so a door closes with the sound at the right frame, speech matches lip movement natively, and ambience matches the environment. In 1.1 you can direct the soundtrack in your prompt: describe dialogue and its emotional tone, or specify ambient cues like "needle crackle," "distant festival music," or "wind rising." Audio is included in the per-second price with no separate charge.

All three generate audio-enabled video. Veo 3.1 (57.5 credits/s with audio) has the best per-frame polish but only 2 aspect ratios and an 8-second cap. Gemini Omni Flash (22 credits/s) has strong physics grounding but caps at 720p and 10 seconds. Happy Horse 1.1 (20-26 credits/s) offers the longest clips of the three (15s), true 1080p, 9 aspect ratios, and the strongest native lip-sync for dialogue — making it the best pick for character performance and multi-format delivery with sound.

Like Happy Horse 1.0, version 1.1 responds best to clean, directive prompts: one subject, one action, one setting, one camera cue — and now one or two sound cues. Prompts can run up to 2,500 characters, but concise direction usually outperforms long descriptions. For dialogue, write the line and describe the delivery: "she whispers, hesitant: 'are you sure?'" The model adapts pacing and tone from that context.

Beyond the standard 16:9, 9:16, and 1:1, you get 4:3 and 3:4 for classic and portrait framing, 5:4 and 4:5 for feed-optimized formats, and the extremes: 21:9 for cinematic ultra-wide and 9:21 for ultra-tall vertical experiences. This means one model covers a full campaign's format matrix — the same scene delivered as a cinema cut, a story, and a feed post without cropping.

Yes. The Animate toggle switches to image-to-video, which animates a first-frame image with the same audio generation. A separate Happy Horse 1.1 Reference variant accepts up to 4 reference images for subject-locked generation — useful for keeping characters and products consistent across a series of clips.

Yes. All Morphed generations — including the synchronized audio and lip-synced speech — are fully commercially licensed for advertising, social media, e-commerce, and broadcast use with no attribution required.