HappyHorse 1.1: 9 reference images, video edit, 1080P, pricing

HappyHorse 1.1 (happyhorse-1.1) is Alibaba’s video model, served on Playcut through DashScope International. Pick it when you need many reference images — it accepts up to 9, the widest reference cap of any video model on Playcut — or when you want longer clips (up to 15 seconds in a single generation, versus Veo’s hard 8-second ceiling). It also generates spoken audio natively from quoted dialogue in your prompt, with no separate text-to-speech pass, and it can edit an existing video — style transfer or subject/object replacement guided by up to 5 reference images. Studio pricing is 56 credits/second at 720p and 72 credits/second at 1080p, so a 10-second 720p clip costs 560 credits.

Specs at a glance

Model id	happyhorse-1.1
Provider	Alibaba (DashScope International)
Modes	text-to-video, image-to-video, reference-to-video, video edit
Aspect ratios (Playcut)	16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 21:9, 9:21 — the provider’s full set
Resolutions	720p, 1080p
Duration	3–15 seconds (integer); video edit has no duration control — output matches the source, capped at 15 s
Max reference images	9 (reference-to-video) · 5 (video edit) · 1 first frame (image-to-video)
Credits	56 cr/s @ 720p · 72 cr/s @ 1080p (video edit bills source + output seconds)
Native speech	Yes — from quoted dialogue in the prompt
Audio input	Not supported
Negative prompts	Not supported (rejected with a 400)

How it compares

Model	Max duration	Max refs	Max resolution	Credits/s @ 720p
HappyHorse 1.1	15 s	9	1080p	56
Veo 3.1	8 s (per call)	3	4k	160
Seedance 2.0	15 s	7	4K	21
Imagine Video (classic)	15 s	7	720p	20
Wan 2.7	15 s	5	1080p	40

Modes

Text-to-video

Tool: generate-video-from-text. Needs only a prompt. Accepts aspectRatio, resolution, durationSeconds, and an optional integer seed.

Note

The provider’s published 3–15 second duration range is documented for HappyHorse 1.0 text-to-video only; Alibaba publishes no range for 1.1 text-to-video. The same 3–15 bounds were verified against the live 1.1 API (the provider rejects out-of-range values with “between 3 and 15 seconds”).

Image-to-video

Tool: generate-video-from-image. Needs a prompt and one imageAssetId — the still becomes the first frame of the clip.

This mode has no aspect-ratio parameter. The output aspect always matches your source image. Sending an explicit aspectRatio other than the 16:9 default is rejected with a 400 rather than silently discarded.

Reference-to-video

Tool: generate-video-from-reference. Needs a prompt and 1–9 referenceImageAssetIds.

References are addressed positionally in the prompt as [Image 1], [Image 2], and so on, in the same order you supply them. Describe the subject alongside the token — the woman in the red qipao in [Image 1] holds identity far better than a bare [Image 1].

Reference image requirements: JPEG, JPG, PNG or WEBP; shortest side at least 400 px; 20 MB maximum per image.

Prompts are capped at 5,000 non-Chinese characters (2,500 Chinese) and are silently truncated beyond that.

Video edit

Tool: generate-video-edit with model: "happyhorse-1.1". Needs a prompt, a source videoAssetId, and optionally up to 5 referenceImageAssetIds for style or subject swaps — for example, “Make the character wear the striped sweater from the image.”

HappyHorse’s editing model ranks #1 on the Artificial Analysis video-editing leaderboard (blind human-preference arena; ahead of Wan 2.7 at #2, as of July 2026).

• Source video: MP4 or MOV, 3–60 seconds, up to 100 MB, above 8 fps.
• Output duration is not a parameter — the output matches the source, capped at 15 seconds. Sources longer than 15 seconds are edited from their first 15 seconds only.
• Audio: audioSetting: "auto" (default — the model decides) or "origin" (keep the source video’s audio).
• Billing: source seconds + output seconds at the resolution’s per-second rate. Playcut holds credits for the maximum up front and automatically refunds the difference once the provider reports the exact billed duration.

Pricing

Resolution	Credits per second	5 s clip	10 s clip	15 s clip
720p	56	280 cr	560 cr	840 cr
1080p	72	360 cr	720 cr	1,080 cr

HappyHorse 1.1 also powers the plus engine tier of the Actor Act pipeline, priced separately at 30 cr/s at 720p or 38 cr/s at 1080p, plus a flat 15 cr scene charge.

Pricing basis

Alibaba publishes no HappyHorse rates on its public pricing pages — the authoritative source is the DashScope Model Studio console, re-verified 2026-07-24 for all three generation modes. Playcut credit prices above are firm and are what you are charged.

Limitations & gotchas

• No audio input. HappyHorse speaks from your prompt text, not from an audio file. Custom voice IDs and audio/clip references are rejected. If you need lip-sync driven by a real audio track, use Wan 2.7 instead.
• Negative prompts are rejected. negativePrompt is not a field on any HappyHorse mode — Playcut returns a 400 instead of silently ignoring it. Fold exclusions into the prompt itself.
• Image-to-video has no aspect ratio. A non-default aspectRatio is rejected with a 400 — the output always follows the source frame. See the mode section above.
• No interpolation, no extension. Both are explicitly rejected with a 400. Video editing runs on the HappyHorse 1.0 edit model upstream (there is no 1.1 edit variant) — same happyhorse-1.1 model id on Playcut.
• Provider default resolution is 1080P. Playcut always sends the resolution explicitly and defaults to 720p, so you are never silently upcharged — but if you omit resolution expecting the cheaper tier from the raw provider API, you would get the expensive one.
• Watermarking is off. The provider stamps a “Happy Horse” badge by default; Playcut disables it on every call.
• Output pixel dimensions are undocumented. Alibaba does not publish exact pixel sizes per aspect ratio at 720P/1080P. Playcut will not guess them.
• Reference ordering is load-bearing. [Image N] tokens must match the order of referenceImageAssetIds.
• Rate limits. DashScope allows roughly 10 requests/second and 5 concurrent jobs per HappyHorse model. Long clips can queue.

How to generate

Set model to happyhorse-1.1 on any of the three video tools. Full request flow (session → submit → wait → download) is in the Quickstart.

Text-to-video

curl -X POST https://api.playcut.ai/api/v1/tools/generate-video-from-text/execute \
  -H "Authorization: Bearer $PLAYCUT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workerId": "65f3a1b2c3d4e5f600000001",
    "model": "happyhorse-1.1",
    "prompt": "A barista slides a latte across the counter, saying: \"Careful, it is hot.\"",
    "aspectRatio": "16:9",
    "resolution": "720p",
    "durationSeconds": 8
  }'

Image-to-video

curl -X POST https://api.playcut.ai/api/v1/tools/generate-video-from-image/execute \
  -H "Authorization: Bearer $PLAYCUT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workerId": "65f3a1b2c3d4e5f600000001",
    "model": "happyhorse-1.1",
    "imageAssetId": "65f3a1b2c3d4e5f600000003",
    "prompt": "The camera pushes in slowly as she turns toward the window",
    "resolution": "1080p",
    "durationSeconds": 6
  }'

Reference-to-video

curl -X POST https://api.playcut.ai/api/v1/tools/generate-video-from-reference/execute \
  -H "Authorization: Bearer $PLAYCUT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workerId": "65f3a1b2c3d4e5f600000001",
    "model": "happyhorse-1.1",
    "referenceImageAssetIds": [
      "65f3a1b2c3d4e5f600000003",
      "65f3a1b2c3d4e5f600000004"
    ],
    "prompt": "The man in the navy jacket from [Image 1] walks into the neon-lit alley from [Image 2]",
    "aspectRatio": "21:9",
    "resolution": "720p",
    "durationSeconds": 12
  }'

From MCP, the equivalent tools are generate-video-from-text, generate-video-from-image, and generate-video-from-reference — same parameters. See the MCP install guide.

Related

• All models — the full catalog index, side by side
• Wan 2.7 — the audio-driven alternative on the same provider
• Seedance 2.0 — cheaper 15-second clips, 7 references
• Veo 3.1 — highest fidelity, 4k, but capped at 8 seconds per call
• Aspect ratios by model
• Resolutions and pricing