Plan connected scenes with multi-shot generation, native audio, visual references, and frame guidance for your next video brief.
Plan a video with multiple shots and a clear sequence of events. Describe what each shot contributes, from establishing a setting to showing a reaction, so the requested structure serves the story rather than collecting unrelated scenes.
Include dialogue, music, and sound effects in the generation brief. Native audio gives sound a part in the scene’s development. Review the full output for wording, timing, and how the soundtrack fits the visual sequence.
In reference mode, combine images, video and audio to guide a scene. Model Studio accepts up to 10 reference images, 5 video clips and 5 audio clips. First/last-frame inputs belong to a separate mode and cannot be mixed with reference inputs.
Guide a generation with a first frame, or define both first and last frames when the task supports it. Pair the visual anchors with a description of the action between them so the transition has a clear creative direction.
Provide a reference video and describe an edit or a continuation. Model Studio supports editing as well as forward, backward and bidirectional extension. Video references share a 15-second total input limit; requests containing video also have a combined input/output duration limit.
| Dimension | Wan 3.0 Prime | Veo 3.1 | HappyHorse 1.1 |
|---|---|---|---|
| Input formats | Text, image, video, audio; task-specific combinations | Text and image; visual references | Text, image, multi-image reference modes |
| Core direction | Multi-shot video with multimodal references | Video generation with audio and visual guidance | Separate text, image, and reference-to-video modes |
| Audio capabilities | Native dialogue, music, and effects | Native audio generation | Audio-supported T2V, I2V and R2V |
| Video duration limits | Model Studio: 2–30s; video inputs have a combined duration limit | Gemini API: 4, 6, or 8s initially; references and 1080p/4K require 8s | Model Studio: 3–15 seconds in T2V, I2V and R2V |
| Output resolution | Model Studio: 480P, 720P, 1080P | Gemini API: 720p; 1080p/4K at 8s; extension at 720p | Model Studio: 480P, 720P, 1080P in all three modes |
| Suggested uses | Multi-shot stories and reference-led scene development | Scenes planned around sound, visual references, and continuation | Character interactions planned from text, images, or references |
Specifications checked October 19, 2026. Sizes and controls refer to the named provider and task. Sources: Wan Prime model guide · Wan 3.0 generation API · Wan 3.0 generation guide · Google Veo 3.1 API guide · Model Studio video model specifications · HappyHorse first-frame API · HappyHorse reference-image API
Break a story into an opening, development and conclusion, then define the subject and action in each shot. This suits short stories told across several shots, with close attention to continuity between them.
Organize reference material and explain how each item should guide the subject, setting or creative direction. Add the intended sound cues to develop an established visual idea into an audiovisual scene.
Use keyframes to establish the intended visual states, then plan an edit or extension. This suits transitions and follow-on actions built around existing material; review joins, subjects and motion for continuity.
Explore video workflows for connected stories, reference-led scenes, and visual transitions.
Yes, in reference mode. Wan Prime accepts combined reference_image, reference_video and reference_audio inputs. First/last-frame mode is separate: its frame inputs cannot be mixed with these reference types.
Model Studio lists 480P, 720P and 1080P, with 2–30-second output and a smart-duration option. When a request includes video, the combined input and output duration must fit the 30-second limit.
Supply reference_video input with editing or continuation instructions. Up to 5 video references are accepted, totaling no more than 15 seconds, with each file at most 100 MB. Forward, backward and bidirectional extension are documented.
No measured speed comparison is included. Alibaba Cloud Model Studio positions Prime as an accelerated Wan 3.0 variant. Use that as provider positioning, then evaluate the exact task, settings, and review workload before basing a production deadline on expected turnaround.
Native dialogue support does not guarantee the wording of every output. Listen to important lines and verify who speaks them, especially when a phrase carries the scene’s message. Check that each spoken line and visible action communicate the same intended event.
