HappyHorse 1.1 References to Video: Full Model Guide
HappyHorse 1.1 reference-to-video (happyhorse-1.1-r2v) generates a brand-new video in which the subject from your reference images — a character, person, product or art style — stays consistent in every frame. Upload up to 9 references, describe the scene in up to 2,500 characters, and the model does the rest. It powers the txtovid reference images to video generator.
Specifications at a glance
- Model ID: happyhorse-1.1-r2v (reference-to-video)
- Image input: up to 9 reference images — JPEG, JPG or PNG, max 20 MB each
- Image size: shorter side > 400 px; 720p or higher recommended
- Aspect ratio: keep references consistent and close to the target video ratio
- Prompt: up to 2,500 characters of free-form scene description
- Output: 720P ($0.14/s list) and 1080P ($0.18/s list) + flat $0.05/s platform fee
- Formats: 16:9, 9:16 and 1:1 — durations of 3, 5, 8, 10 or 15 seconds
Identity lock, explained
Text-to-video invents everything and image-to-video freezes your composition — reference-to-video sits in between: the model extracts the identity of the subject from your reference images and re-stages it in whatever scene your prompt describes. Camera, setting, lighting and aspect ratio remain fully promptable.
That's what makes r2v the tool for series content: the same mascot in ten different ads, the same character across an episodic story, the same product in every seasonal campaign — without the identity drift that plagues pure text-to-video.
Getting the best results
Feed the model 3–9 sharp images of the same subject from different angles and poses, shot in similar quality. Then use the long prompt budget: name the subject (“the character from the references”), and describe action, setting, camera and lighting in detail. Keep one hero subject per run — multi-character scenes work better as separate generations.