HappyHorse 1.0 Video Edit
What is HappyHorse 1.0 Video Edit?
HappyHorse 1.0 Video Edit is the video-to-video editing endpoint in the HappyHorse 1.0 family. Send a source video and a prompt describing the change, optionally add up to 5 reference images for style hints, and the model returns an edited clip at 720p or 1080p. It is a separate model id (alibaba/happyhorse-1.0-video-edit) on the same task endpoint as base HappyHorse 1.0, so you keep the same auth, polling, and webhook flow. The audio_setting field decides whether the original audio is kept (origin) or rewritten alongside the visual edit (auto), which makes the same endpoint usable for clean style transfers and full reinterpretations.
Key features of HappyHorse 1.0 Video Edit
Five features cover the controls you'll touch in production.
Video-to-video edits from a single prompt
Send a source video and an instruction in plain language. The model rewrites the look, mood, or composition of the clip without you mapping the change to per-frame parameters.
Up to 5 reference images for style guidance
Drop in moodboard stills and the model lifts palette, lighting, and art-direction cues from them. Useful when the change is "make it look like this" rather than "do this specific thing".
Original-audio preservation on demand
Set `audio_setting` to `origin` and the source soundtrack passes through untouched. Set it to `auto` and the model decides, which is the right call for full reinterpretations where the new look implies new sound.
720p or 1080p output
Same resolution presets as base HappyHorse 1.0. Default is 1080p for delivery; switch to 720p when you want a faster pass on a long clip.
Seed control for reproducible edits
Pass an integer seed (0 to 2,147,483,647) and the same source video, prompt, references, and seed return the same edit. Useful for client review rounds and locking a take you want to keep.
Best for
Style transfer for social cuts
Apply a moodboard palette across a brand shoot for a platform-specific repurpose. Up to 5 reference images carry the look without per-frame rotoscoping.
Late-stage creative pickups
Client wants a different prop or a warmer grade. One prompt, one call, and the source clip comes back edited at 720p or 1080p.
Audio-preserving look changes
Set `audio_setting` to `origin` and the original voiceover or dialogue stays intact while the picture is restyled.
Brand adaptation at scale
Run a backlog of catalogue clips through the same prompt and reference set so a seasonal launch goes out with consistent look and pacing.
Post-production passes on existing footage
Use it for finishing rather than generation: noise, grade, and atmosphere changes on shots you already shot.
Iteration with locked seeds
Lock a winning edit with a seed, then change the prompt one variable at a time without losing the rest of the look.
Use cases
Re-grade a hero brand spot for a moody seasonal launch by sending the original clip with a one-line prompt ("warmer golden-hour palette, slightly desaturated") and three moodboard stills as references; original audio survives via audio_setting: origin. Repurpose a vertical product clip into an "anime cel" cut for a platform-specific campaign with a single reference frame. Fix a late client note (cup becomes a wine glass, sky becomes sunset) without re-rendering from scratch. Chain edits across a backlog of catalogue clips by reusing the same prompt and seed across requests so the look stays consistent.
API examples
Call HappyHorse 1.0 Video Edit from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/video/alibaba/happyhorse-1.0-video-edit.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "alibaba/happyhorse-1.0-video-edit",
"input": {
"prompt": "Convert the scene to a watercolor painting style",
"video_url": "https://example.com/scene.mp4",
"resolution": "1080P"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished video URL.
FAQs
People also ask
HappyHorse 1.0 Video Edit is a video-to-video editing endpoint built on the HappyHorse 1.0 architecture. Pass a source video and a text prompt describing the change, optionally add up to 5 reference images for style guidance, and get a styled output back at 720p or 1080p.
Base HappyHorse 1.0 generates new video from a prompt, a first frame, or character references. Video Edit takes an existing video as input and rewrites it according to the prompt. It is a separate model id (alibaba/happyhorse-1.0-video-edit) on the same endpoint.
Up to 5 per call. The model uses them as style hints, so you can carry palette, lighting, or art-direction cues from a moodboard onto the source clip without describing each detail in prose.
Yes. Set audio_setting to origin and the original audio is preserved end-to-end. Leave it on auto and the model decides how to handle audio in the output, which fits style transfers where the new look implies new sound.
720p or 1080p, matching base HappyHorse 1.0. The default is 1080p; switch to 720p when you want a faster turnaround on a long clip.
Up to 5,000 non-Chinese characters or 2,500 Chinese characters per call. Anything past the cap is truncated, so trim long edit notes before sending.
Yes. Pass an integer seed in the range 0 to 2,147,483,647 alongside the prompt and the source video. Re-sending the same inputs and seed returns the same edited clip.