Video GenerationActive
MiniMax H3 Lite R2V Video Generator
byminimax
MiniMax H3 Lite R2V generates 4–15s videos from a prompt plus reference images, clips, and audio. It helps keep characters, style, and motion consistent.
Overview
MiniMax H3 Lite R2V is a reference-to-video model made by MiniMax. You provide a written scene description plus optional reference images, short videos, and audio clips. The model uses those references to guide identity, style, motion, camera feel, and sound. It returns a single short MP4 clip, which helps you keep continuity across takes.
What you can build
- Character-consistent shots from a small set of reference photos
- Style-matched clips for ads, product shots, and branded visuals
- Motion-guided shots using a short driving clip as a movement reference
- Voice- or music-guided clips using an audio reference plus visual references
- Rapid storyboard passes where each shot shares the same cast references
Inputs
- A required text description of the clip you want. If you upload references, you should refer to them by order, like “Image 1”, “Video 1”, and “Audio 1”.
- Optional reference images (JPG, PNG, or WebP). You can upload up to 9 images for subject or style guidance.
- Optional reference video clips (commonly MP4). You can upload up to 3 clips. Each clip must be 2–15 seconds, and the combined reference-video time must be 15 seconds or less.
- Optional reference audio clips (commonly MP3 or WAV). You can upload up to 3 clips. Each clip must be 2–15 seconds, and the combined reference-audio time must be 15 seconds or less. Audio references can’t be used alone, so include at least 1 image or video reference.
- A required target clip length in whole seconds. The allowed range is 4–15 seconds.
- A required output resolution choice. H3 Lite supports 480p and 768p output.
- An optional output aspect ratio selection. Supported ratios include 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16.
Outputs
- One generated MP4 video file.
- The video matches the requested duration (4–15 seconds) and the selected resolution (480p or 768p).
- The file includes a synchronized audio track when sound is generated or guided by references.
Limitations
- The model only generates short clips. You can’t request longer than 15 seconds in one run.
- Maximum output resolution is 768p on H3 Lite.
- You must pick a fixed aspect ratio from the supported list.
- Reference inputs have hard caps: up to 9 images, up to 3 video clips, up to 3 audio clips, and no more than 12 total reference files.
- Audio references require at least one visual reference. Audio-only requests aren’t supported for R2V.
- Low-quality or inconsistent references can reduce control. Blurry faces, heavy compression, watermarks, and fast-cut video references can weaken identity and motion transfer.
Safety & compliance
- Only use references you have rights to use. Get consent before using a real person’s likeness or voice.
- Don’t use the model to generate deceptive impersonations or to mislead viewers.
- Follow MiniMax policies and any applicable platform rules for prohibited content and disclosure requirements.
Example prompts
Great starting points for h3 lite r2v.
Image 1 is the brass pocket watch. An elderly clockmaker with silver hair and round glasses sits at a wooden workbench in his cozy clock repair shop, picks up the pocket watch, gently winds its crown, then opens the engraved lid and smiles as the second hand starts ticking. Warm lamp light, ticking wall clocks behind him, slow push-in.Video Generation
Image 1 is the wooden violin and bow. A young street musician with dark curly hair and a gray wool coat stands under a stone archway on a cobblestone street at dusk, raises the violin to his chin and plays it, the bow gliding smoothly across the strings as he sways gently with the music, a couple walking by slows down to listen. Warm street lamp light, slow cinematic push-in.Video Generation
Image 1 is the gray metal food truck. The food truck is parked on a busy Manhattan street at lunchtime, a vendor in an apron leans out of the serving window and hands a wrapped hot dog to a smiling customer, a short line of office workers waits on the sidewalk and yellow taxis drive past behind. Bright midday light, street-level camera slowly moving sideways.Video Generation
Image 1 is the straw sun hat. A woman in a white linen dress walks along a sunny beach wearing the straw sun hat, a gust of wind lifts the hat off her head and it tumbles across the sand, she laughs and runs after it and catches it by the ribbon. Bright summer light, waves rolling in, camera tracking alongside her.Video Generation
API quick start
Run h3 lite r2v with a single API call.
{
"prompt": "Image 1 is the brass pocket watch. An eld…",
"inputImage": "https://your-cdn.com/input.png",
"duration": 5,
"resolution": "768P"
}curl -X POST "https://api.wiro.ai/v1/Run/minimax/h3-lite-r2v" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "Image 1 is the brass pocket watch. An eld…",
"inputImage": "https://your-cdn.com/input.png",
"duration": 5,
"resolution": "768P"
}
JSON