minimax-h3-max/reference-to-video
MiniMax H3 Max Reference to video: multimodal refs to 480p/768p clips
MiniMax H3 Fast bills every resolution below at the same rate. Generation is faster than standard MiniMax H3, and detailed prompt instructions are followed less strictly.
| Parameter | Required | Type | Default | Range / Options | Description |
|---|---|---|---|---|---|
| image* | Yes (*) | Image | None | JPEG, PNG, WebP, and similar; URL up to 30 MB | Opening frame of the shot. |
| last_image* | Yes (*) | Image | None | Same limits as image | Frame the clip should land on. |
| prompt* | Yes (*) | String | Example prompt | 1 or more characters | Describe the motion between the two frames, plus camera and sound. |
| aspect_ratio | No | String | 16:9 | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | Output frame shape. |
| resolution | No | String | 768p | 480p, 768p, 1080p, 2k | All four sizes bill at the same per-second rate. |
| duration | No | Integer | 5 | 5-15 seconds | Longer clips give the transition more time, and cost more. |
MiniMax H3 Fast is $0.02 per second of generated video at 480p, 768p, 1080p, and 2K. Reference images and a reference video do not add a separate charge on the reference page.
| Duration | Price |
|---|---|
| 5 seconds | $0.10 |
| 10 seconds | $0.20 |
| 15 seconds | $0.30 |
A batch multiplies that total by the number of outputs.
MiniMax H3 Max Reference to video: multimodal refs to 480p/768p clips
Seedance 2.5 4K Text to Video: prompt to 4K cinematic clips
LTX 2.5 Pro text-to-video with synced AV in quality mode
Create lifelike speech-synced visuals from scripts or clips with Kling Lipsync for precise facial animation and realistic results.
Consistent characters, objects, and scenes in any setting or angle.
Turn static visuals into smooth motion with Hailuo 2.3 for rapid, realistic video creation.
MiniMax H3 Fast builds a 5-15 second video with native audio between an opening image and a closing image. It is useful for transitions, product reveals, and short story beats when both ends of the shot are already designed.
MiniMax H3 Fast is faster than standard MiniMax H3 and follows detailed instructions less strictly. Describe one clear change between the two frames. Extra style notes and on-screen text are easier for MiniMax H3 Fast to miss.
MiniMax H3 Fast requires a prompt, an opening image, and a closing image. Images can be JPEG, PNG, or WebP, and a URL image may be up to 30 MB. Duration is 5-15 seconds at 480p, 768p, 1080p, or 2K.
MiniMax H3 Fast uses the two images as the start and end of the motion, but it does not guarantee a pixel-locked match. Leave more duration when the two frames differ a lot.
Yes. MiniMax H3 Fast writes audio with the picture. Put the intended sound in the prompt, and treat the result as a faster, less literal pass than standard MiniMax H3.
Use MiniMax H3 Fast for a quicker, cheaper bridge between two frames. Use standard MiniMax H3 when the path between those frames has to follow a specific prompt closely.
MiniMax H3 Fast costs $0.02 per second of generated video at 480p, 768p, 1080p, and 2K. A 5-second clip is $0.10, a 10-second clip is $0.20, and a 15-second clip is $0.30. Reference images and a reference video do not add a separate charge. Generations spend USD credits, and new accounts usually receive a trial balance.
Yes. Build the shot in the Playground Web UI, then send the same MiniMax H3 Fast model and parameters over the RunComfy HTTP API. The page does not promise a fixed queue time or unlimited concurrency.
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.
