HappyHorse Video Model Family Comparison: How to Choose Across Four Core Versions

II. Four Models Deep Dive

Model 1: v1-text-to-video

This is HappyHorse's base text-to-video model for direct prompt-driven generation.

{
  "prompt": "Text prompt for scene/style (max 5000 chars)",
  "resolution": "720p / 1080p",
  "aspect_ratio": "16:9 / 9:16 / 1:1 / 4:3 / 3:4",
  "duration": 5,
  "seed": "Optional (0-2147483647)"
}

Key points: duration supports 3-15s, five aspect ratios, bilingual prompts up to 5000 characters.

Use cases: concept videos, creative ads, product demos, atmosphere clips.

Model 2: v1-image-to-video

This model generates motion from one first-frame image plus an optional prompt.

{
  "prompt": "Optional text prompt for constraints (max 5000 chars)",
  "image_urls": ["First-frame image URL (required, exactly one)"],
  "resolution": "720p / 1080p",
  "duration": 5,
  "seed": "Optional"
}

Image limits: JPEG/JPG/PNG/WEBP; width and height >= 300px; ratio 1:2.5 to 2.5:1; file <= 10MB.

Key points: prompt is optional; without prompt, motion is driven mainly by image content; duration 3-15s.

Use cases: animate old photos, illustration animation, dynamic product visuals.

Model 3: v1-reference-to-video (Signature Feature)

This is HappyHorse's signature model for multi-character or multi-object consistency based on ordered reference images.

{
  "prompt": "Use character1/character2/... to refer to subjects (max 5000 chars)",
  "image_urls": ["Reference images mapped to character1, character2..."],
  "resolution": "720p / 1080p",
  "aspect_ratio": "16:9 / 9:16 / 1:1 / 4:3 / 3:4",
  "duration": 5,
  "seed": "Optional"
}

Core mechanism: 1-9 images; order matters; prompt uses character1/character2 references; model keeps appearance consistency.

Image requirements: JPEG/JPG/PNG/WEBP; short side >= 400px (720p+ recommended); avoid blurry or over-compressed images; <= 10MB.

Use cases: character-consistent videos, multi-character ads, IP animation.

Model 4: v1-video-edit (Signature Feature)

This is the dedicated HappyHorse editing model with up to 60-second edit length and natural-language instruction control.

{
  "prompt": "Natural language edit instruction (max 5000 chars)",
  "video_url": "Input video URL (required, one video)",
  "reference_image": "Optional reference image URLs (0-5)",
  "resolution": "720p / 1080p",
  "audio_setting": "auto / origin",
  "seed": "Optional"
}

Video limits: MP4/MOV (H.264 recommended); 3-60s; long side <= 2160px; short side >= 320px; ratio 1:2.5 to 2.5:1; file <= 100MB; fps > 8.

Reference limits: 0-5 images; JPEG/JPG/PNG/WEBP; width and height >= 300px; ratio 1:2.5 to 2.5:1; <= 10MB.

Use cases: outfit swap, background replacement, style transfer, local edits.