wan-2-6/video-to-video
Transforms reference clips into 1080p short videos with precise motion and voice alignment.
This page runs the open-weights reference path in the MiniMax H3 family. Attach the media you want the model to follow, describe the scene, and MiniMax H3 Open Reference To Video returns an MP4 with native stereo audio at 480p or 768p.
The idea is simple: images pin what things look like, videos pin how things move, and audio guides the sound. Refer to each input in your prompt and say what job it does, so the model carries a subject, style, or motion from shot to shot instead of guessing.
1) Attach your references — images for identity or style, clips for motion, optional audio for a matched voice or track.
2) Write a timeline prompt with timestamped beats, and name what each reference supplies.
3) Set the camera language explicitly, including when to hold.
4) Add an Audio: line for entrance cues and ambience.
5) Pick aspect ratio, resolution, and duration, then generate and refine one reference or clause at a time.
For the higher-resolution MiniMax H3 reference-to-video tier, see the related MiniMax H3 models on RunComfy.
Transforms reference clips into 1080p short videos with precise motion and voice alignment.
Cinematic 4K image-to-video at $0.42 per second of output.
Transforms static characters into smooth motion clips for flexible creative workflows
Seedance 2.5: Animate a still image into cinematic AI video
HappyHorse 1.0 with native 1080p output, cinematic motion, and multi-shot consistency.
Seedance 2.5 480p: Fast, low-cost still-to-video animation drafts
MiniMax H3 Open Reference To Video builds a short clip from a prompt plus reference media, so a subject, product, style, or motion stays consistent across shots. It fits character continuity, product reveals, and style matching where the look must follow real references instead of prompt wording alone.
MiniMax H3 Open Reference To Video accepts up to 9 reference images, 3 reference videos, and 3 reference audios. At least one reference image or video is required, and audio cannot be provided on its own. Name each reference in the prompt so the model knows what it supplies.
MiniMax H3 Open Reference To Video is the open-weights path with 480p and 768p output. The official MiniMax H3 reference-to-video model on RunComfy uses a different resolution ladder such as 768p and 2K. Choose the open-weights route for the tiers and pricing shown on this page.
Yes. MiniMax H3 Open Reference To Video generates native stereo audio together with the picture in a single pass. A reference video's own soundtrack is used automatically, and you can add standalone reference audios or an Audio: line in the prompt to steer ambience, dialogue, or a matched voice.
Write MiniMax H3 Open Reference To Video prompts as a timeline with timestamped beats, and assign a job to each reference: which supplies identity, which supplies style, and which supplies motion. State the camera language explicitly, including when to hold, and keep readable on-screen text in quotes.
Yes. Prototype in the RunComfy model UI, then call the same MiniMax H3 Open Reference To Video template through the RunComfy API with identical parameters. Submission returns a task id; poll until it completes and read the output video. Both paths consume usd / credits.
MiniMax H3 Open Reference To Video is billed per generated second—$0.06 per second at 480p and $0.11 per second at 768p—plus per-reference charges: $0.03 per reference image, $0.03 per reference audio, and $0.06 per second of reference video. For example, a 10s 480p clip with 2 reference images and a 5s reference video costs 10 × $0.06 + 2 × $0.03 + 5 × $0.06 = $0.96.
MiniMax H3 Open Reference To Video outputs 480p or 768p at 24 fps, with durations from 5 to 15 seconds and seven aspect ratios. Reference videos are supported at 480p output only and share a 15-second total budget; standalone reference audios are trimmed to 15 seconds each.
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.
