HappyHorse 1.1 Reference-to-Video & Multi-Shot Short Drama Workflow Guide
If you use HappyHorse for short-drama storyboards, ad previsualization, or serialized character content, the biggest 1.1 upgrade is Reference-to-Video plus multi-shot, multi-asset fusion understanding. Unlike pure text-to-video where you “roll once and hope,” HappyHorse 1.1 lets you feed up to 9 character or product reference images and keep the subject consistent across a 3–15 second clip with more controllable shot continuity.
This HappyHorse tutorial focuses on “reference-to-video + multi-shot short drama workflow”—a different theme from past case studies and the 1.0 product intro. You will learn how to combine text-to-video, image-to-video, and reference-to-video in the HappyHorse video workspace, write reusable multi-shot prompts, and balance cost vs. quality between 720P and 1080P.
1. Why HappyHorse 1.1 Fits Multi-Shot Short Drama
Short drama, ad storyboards, and serialized IP content share one pain point: faces change, outfits drift, product details wander. HappyHorse 1.1 targets that scenario directly:
| Capability | Value for multi-shot creation |
|---|---|
| Reference-to-video (up to 9 refs) | Lock protagonist face, outfit, and product look across shots |
| Stronger instruction following | Timeline and camera moves in complex prompts execute more reliably |
| Better motion | Smoother transitions, tracking, push/pull—less “slideshow” feel |
| Native audio–video sync | Dialogue, VO, and lip sync in one pass for talking-head shots |
| 3–15 s × any aspect ratio | Vertical short drama, horizontal ads, 1:1 product showcases |
HappyHorse ranking #1 on both Artificial Analysis video leaderboards reflects its strength on complex prompts and motion quality. For creators who need fast storyboard validation, HappyHorse 1.1 is a cost-effective AI video generation model choice.
2. Reference-to-Video Core Capabilities
Reference-to-video is the HappyHorse mode most tied to subject consistency. Think of it as: tell the model what the hero, product packaging, and scene style look like—then let it fill in motion.
Best use cases
- Short-drama boards: Same lead in elevator, hallway, and room with stable face and hair
- E-commerce ads: Same SKU at different angles with logo and colors intact
- Game/anime previz: Character turnaround + action description → quick motion sample
- Brand series: Fixed VI colors and mascot design across multiple 5 s bumpers
How it works with image-to-video and text-to-video
| Mode | Typical input | Role in a storyboard pipeline |
|---|---|---|
| Text-to-video | Text prompt only | Establishing shots, mood, transitions without a fixed subject |
| Image-to-video | First frame + optional text | Single-shot motion from a locked composition |
| Reference-to-video | Multiple refs + text | Core tool for cross-shot subject consistency |
| Video editing | Existing video | Grade, local repaint, style unification |
In the HappyHorse workspace, lock the subject with reference-to-video first, add transitions via image/text-to-video, then use video editing to unify color if needed.
3. Workspace Prep: Assets, Aspect Ratio, Parameters
Before opening the HappyHorse video workspace (app.happy-horse.xin), prepare:
1. Reference image checklist (3–9 images)
- Character: front bust, 3/4 profile, full-body standing (1 each; JPEG/PNG; clear subject)
- Wardrobe/props: 1–2 close-ups (avoid heavy filters)
- Scene/lighting: 1–2 mood references (need not match final framing)
- Product (e-commerce): front, 45°, logo detail (1 each)
HappyHorse 1.1 fuses multiple references, but quality beats quantity: blurry, watermarked, or crowded group shots hurt consistency.
2. Aspect ratio
| Ratio | Common use | HappyHorse note |
|---|---|---|
| 9:16 | TikTok/Reels short drama, feed ads | Default for vertical boards |
| 16:9 | Brand TVC, YouTube | Horizontal narrative, multi-person frames |
| 1:1 | E-commerce hero video, social square | Centered product |
| Custom | Special ad sizes | HappyHorse supports arbitrary aspect ratios |
3. Resolution and duration
- 720P: Prompt trials, motion and lip-sync checks—lower cost
- 1080P: Final delivery and external campaigns
- Duration: 3–15 s per generation; multi-shot drama is often several 5–8 s clips edited together
4. Five-Step Workflow: Storyboard Script to HappyHorse Output
A reusable HappyHorse multi-shot workflow for teams:
Step 1: Write a shot list (Excel/Notion)
One row per shot: Shot ID | Duration | Framing | Action/dialogue | Ref IDs | Mode
Example:
| Shot | Duration | Mode | Notes |
|---|---|---|---|
| S01 | 3s | Text-to-video | City night establishing |
| S02 | 5s | Reference-to-video | Lead exits elevator, refs R1–R4 |
| S03 | 4s | Reference-to-video | Lead enters room, refs R1–R5 |
| S04 | 3s | Image-to-video | First frame = last frame of S03, door close CU |
Step 2: Upload and label references
Upload R1–R9 in the HappyHorse workspace with numbered filenames (e.g. hero-front.jpg). In prompts, say “same female lead as reference R1.”
Step 3: Generate shot by shot
Do not cram a full 15 s story into one prompt unless you are experimenting with one-take generation. HappyHorse executes single clear shot goals more reliably.
Step 4: Iterate at 720P
Run 2–3 versions per shot at 720P, 3–5 s; pick the take with best motion and lip sync.
Step 5: 1080P finals + edit
Regenerate winners at 1080P; assemble in CapCut, Premiere, etc., with subtitles and BGM.
5. Strategy for 9 Reference Images
HappyHorse 1.1 accepts up to 9 references. Layer by priority:
- R1–R3 (required): face, hairstyle, main outfit color
- R4–R5 (strongly recommended): full-body proportions, signature accessories
- R6–R7 (optional): scene lighting, brand palette
- R8–R9 (optional): prop/product detail, second character
Bad: nine unrelated web angles with no clear subject.
Good: six hero + two scene + one product; open prompt with “same woman as refs R1–R6.”
6. Multi-Shot Prompt Structure and Timeline Writing
HappyHorse understands prompts with time segments well. Recommended structure:
【Overall】15 s vertical short-drama board, cinematic realism, same heroine as refs R1–R5, shallow DOF.
【0–3 s】Medium shot, heroine in elevator, doors open slowly, warm overhead light, slight upward glance.
【3–7 s】Tracking shot, heroine steps into hallway, natural gait, hair sways, camera follows from rear-side.
【7–11 s】Over-shoulder, heroine pushes door; cold interior light through gap, warm/cool contrast.
【11–15 s】Near POV entering room; focus rack from handle to city night outside window.
Dialogue (Chinese): “Finally here.” Lip sync via HappyHorse native audio–video alignment.
Avoid: face swap, outfit color shift, crowd blocking face.
Prompt tips
- Subject → action → camera for stable HappyHorse 1.1 execution
- Use “avoid” lines to reduce common failures (extra fingers, face drift)
- Specify language for dialogue (7 languages with lip alignment supported)
- Break complex camera moves into short sentences; avoid 200+ word paragraphs
7. 720P Trials vs. 1080P Delivery
HappyHorse bills per second; 720P vs. 1080P prices differ. Running everything at 1080P scales cost linearly with shot count.
Recommended strategy:
- Run all boards at 720P → validate story and consistency
- Upgrade only close-ups + dialogue shots to 1080P
- Keep establishing/wide shots at 720P or sharpen lightly in edit
Official architecture cites ~38 s to generate 5 s 1080P on a single H100—queue time extra. Batch storyboards off-peak when possible.
8. Audio–Video Sync and Multilingual Lips
HappyHorse 1.1’s native audio–video sync supports dialogue-heavy short drama:
- Keep one language per series (Chinese dialogue → all Chinese prompts)
- Explicitly request “lip sync with dialogue”
- For global assets, specify English, Japanese, etc., and use 7-language lip alignment to cut dubbing fixes
If lip sync fails, shorten dialogue in the prompt before bumping resolution—long lines in 5 s clips misalign more often.
9. Walkthrough: 15 s Vertical Short “Running Late”
Plot: Lead rushes from subway to office door.
Refs: R1 front, R2 profile, R3 full-body (same OL outfit)
Modes: S01 text-to-video establishing; S02–S04 reference-to-video
S02 prompt example:
Same female lead as refs R1–R3, 9:16 vertical, 5 s. Medium shot in subway car, checks watch, anxious expression, tunnel lights streak past window. Slight handheld camera, realistic style. Dialogue: “I’m late!” Lip sync.
S03 prompt example:
Same lead (refs R1–R3), 5 s. Platform, brisk run in heels, side tracking shot, blurred crowd background. Smooth motion, no limb distortion.
Add SFX transitions + subtitles in edit for a deployable short-drama teaser board.
10. FAQ
Q1: Reference-to-video or image-to-video for HappyHorse?
A: HappyHorse reference-to-video when the same hero/product must persist across shots; image-to-video when animating one locked first frame. Multi-shot drama is usually reference-first.
Q2: Must I use all 9 reference slots?
A: No. 3–5 high-quality face/full-body refs often beat nine noisy images.
Q3: Can one 15 s HappyHorse generation cover the whole board?
A: Split into 3–5 s shots and edit together—much higher success than one long prompt.
Q4: Subject still inconsistent?
A: Check refs are the same person; stress “same protagonist” in prompt; reduce wide shots; use HappyHorse video editing for local repaint.
Q5: Where do new users start?
A: Open the HappyHorse video workspace from the official site—new users get 10 free seconds; prove the shot list at 720P before 1080P finals.
11. Conclusion
HappyHorse 1.1 folds reference-to-video, multi-shot understanding, and audio–video sync into one AI video generation pipeline—turning short-drama boards from “gacha” into a repeatable industrial workflow. With reference layout, timeline prompts, and 720P→1080P iteration, you can ship serialized content steadily from the HappyHorse workspace.
Next step: open the HappyHorse video workspace, use the shot-list template here, and render three 5 s reference-to-video takes—compare subject consistency, then expand to a full 15 s board.
(Based on public HappyHorse 1.1 capabilities; features and pricing follow the live HappyHorse workspace. Published: 2026-06-26.)
Ready to test these prompts? Click the button below to open HappyHorse.
Open HappyHorse Workspace