AIREITER
HappyHorse 1.1 Reference-to-Video Generation
HappyHorse 1.1, also written as Happy Horse 1.1, is the reference-first option on AIReiter. It generates a new video from one to nine ordered images and a prompt that explicitly links each subject, object, or scene to its source image.
One to nine ordered references
Combine people, clothing, products, props, environments, and style references in one ordered image set.
Explicit prompt-to-image links
Use [Image 1], [Image 2], and later markers to tell the model exactly which uploaded image supplies each visual element.
Nine output aspect ratios
Create landscape, portrait, square, standard, ultra-tall, or cinematic framing without changing the reference order.
Playground and API parity
The same prompt, image order, resolution, aspect ratio, duration, and seed controls are available through happyhorse_1_1 API requests.
AIReiter-Tested HappyHorse 1.1 Reference Workflow
This input-and-output pair was generated through the AIReiter test environment on August 20, 2026. The example shows the exact reference image, prompt, request settings, and delivered video.
Input reference: a clear dog portrait
Prompt: The dog in [Image 1] turns its head gently toward the camera, fixed shot, soft daylight.
Request: 1080p, 16:9, 5 seconds, random seed.
The prompt names both the subject and its image marker, then uses one restrained action and one camera instruction.
Output result: reference-driven head movement
Observed task time: about 209 seconds end to end in this test. Queue time can vary.
The output uses the dog from [Image 1] as the subject while adding the requested head turn under a fixed camera.
How to Prepare HappyHorse 1.1 Reference Images
Reference quality and prompt organization matter more as the number of uploaded images increases.
Use a clear subject per image
Prefer sharp references with an obvious person, product, prop, garment, or environment. Avoid images where several similar subjects compete for attention.
Keep proportions compatible
Use references with similar proportions when possible and keep the short-to-long edge ratio at or above 0.4 to reduce framing conflicts.
Assign every image a role
Decide whether each upload supplies a character, wardrobe item, object, style, or scene before writing the prompt.
Name markers in upload order
The first image is [Image 1], the second is [Image 2], and so on. Reordering uploads without updating the prompt can swap identities and props.
Practical HappyHorse 1.1 Use Cases
Character and wardrobe direction
Use separate references for a person, clothing, and accessories, then identify each item with its own [Image N] marker.
Product videos with scene guidance
Combine a product reference with an environment or style reference to guide both the featured object and its presentation.
Prop and environment continuity
Anchor a recurring object and location with separate images so the prompt can describe how they interact in the generated scene.
Multi-reference visual concepts
Assemble a subject, prop, wardrobe item, and setting from different images while keeping their roles explicit in the prompt.
HappyHorse 1.1 Parameters and Reference Limits
Set the output after the reference order and prompt markers are finalized.
Image requirements
Provide one to nine publicly accessible JPEG, JPG, PNG, or WEBP images. Each image must have a short edge of at least 400 pixels; 720p or larger references are recommended.
Resolution
Choose 720p for standard output or 1080p when higher visual detail is required. The selected tier affects price.
Aspect ratio
Choose 16:9, 9:16, 3:4, 4:3, 4:5, 5:4, 1:1, 9:21, or 21:9.
Duration
Choose any whole-second duration from 3 to 15 seconds.
Seed
Use a positive integer from 1 to 2147483647 for more repeatable iterations, or omit it for a random seed.
Prompt length
Prompts support up to 2,500 Chinese characters or 5,000 non-Chinese characters. Put reference assignments and the main action before secondary styling details.
How to Prompt HappyHorse 1.1
Write the prompt as a short production brief that maps every important visual element to an uploaded image.
Upload references in final order
Add the images in the exact order you will use in the prompt. Confirm the sequence before writing detailed instructions.
Identify the source of each element
Write phrases such as the woman in [Image 1], the jacket in [Image 2], and the street in [Image 3] instead of listing markers without roles.
Describe one coherent action
After assigning references, state the main movement and one compatible camera instruction. Avoid asking the same subject to perform conflicting actions.
Review the mapping before submission
Check that every [Image N] marker still matches the same image_url position, then select resolution, aspect ratio, duration, and seed.
HappyHorse 1.1 vs HappyHorse 1.0
Choose by workflow rather than treating 1.1 as a universal replacement for 1.0.
Choose HappyHorse 1.1 for reference-first scenes
Use happyhorse_1_1 when one to nine ordered images define the people, objects, clothing, products, or locations in a new video.
Choose HappyHorse 1.0 for broader generation
Use happyhorse_1_0 for text-to-video, one-image animation, reference-to-video, or source-video editing from one model page.
Need to edit a source video?
Use HappyHorse 1.0. The current HappyHorse 1.1 integration accepts reference images, not a source video.
Need explicit multi-image roles?
Use HappyHorse 1.1 and map every important image to its [Image N] marker before describing the action.
Use HappyHorse 1.1 Through the AIReiter API
Submit one reference-to-video task with ordered images and retrieve the generated result asynchronously.
Model ID
Use happyhorse_1_1 as the model value.
Submit endpoint
POST /api/openapi/submit accepts prompt, image_url, resolution, aspect_ratio, video_length, optional seed, and a unique out_task_id.
Query endpoint
POST /api/openapi/query accepts the same out_task_id. Poll until the task is completed or failed.
Reference syntax
Keep every [Image N] marker aligned with the same position in image_url. Misaligned ordering can confuse characters, props, clothing, and scenes.
Reference count validation
Send between one and nine image URLs. Each URL must be publicly accessible to the provider during generation.
Related AI Video Models
HappyHorse 1.1 FAQ
Answers about reference preparation, prompt markers, output settings, subject confusion, and the difference between HappyHorse 1.0 and 1.1.
/ 01What does HappyHorse 1.1 generate?
The current AIReiter integration provides reference-to-video generation from one to nine ordered images and a required prompt.
/ 02How do I refer to an uploaded image?
Use [Image 1] for the first image_url item, [Image 2] for the second, and continue in order through [Image 9]. Name the role of each image in the same phrase.
/ 03Can I generate from only one reference image?
Yes. Upload one image, refer to its subject as [Image 1], and describe the desired movement, camera behavior, and environment.
/ 04How many reference images can I upload?
You can provide between one and nine images. Use clear images with a short edge of at least 400 pixels and publicly accessible URLs.
/ 05Which durations and resolutions are supported?
Choose any whole-second duration from 3 to 15 seconds and either 720p or 1080p output.
/ 06Why did the model confuse two referenced subjects?
Check that image_url order matches every [Image N] marker, then describe the relevant person, object, clothing, product, or scene explicitly. Avoid using nearly identical unlabeled references.
/ 07What is the difference between HappyHorse 1.0 and 1.1?
HappyHorse 1.1 is reference-to-video only in the current integration. HappyHorse 1.0 also provides text-to-video, image-to-video, reference-to-video, and source-video editing.
/ 08Which API model ID should I use?
Use happyhorse_1_1 with POST /api/openapi/submit, then poll POST /api/openapi/query using the same out_task_id.