wan-2-2/speech-to-video
Turn photos into expressive videos with synced voice motion.
MiniMax H3 Max is MiniMax's post-trained H3 video line for short clips with picture and sound rendered together—and it is built for speed. A 10-second clip often comes back in roughly a dozen seconds, so you can draft, revise, and lock shots in a tight loop.
On RunComfy you can start from a written shot list (text-to-video) or lock the opening frame from a still (image-to-video), then optionally steer the ending with a last frame.
This page runs the image-to-video path. The sibling text-to-video page shares the same MiniMax H3 Max family when you prefer to invent the first frame from words alone. Either way, MiniMax H3 Max aims for strong prompt following without a separate audio pass.
balanced for a quick rewrite or quality when you want a richer expansion before generation.480p or 768p for drafts and delivery-ready previews.| Parameter | Required | Type | Default | Range / Options | Description |
|---|---|---|---|---|---|
| image_url * | Yes (*) | string (URL) | Sample first frame | Image URL | Opening frame that anchors subject, composition, and style. Aspect ratio follows this image. |
| prompt * | Yes (*) | string | Sample motion brief | Text | Motion, camera, lighting, style, and optional audio direction. |
| end_image_url | No | string (URL) | None | Image URL | Optional last frame for first-to-last keyframe generation. |
| duration | No | integer | 5 | 5-15 | Clip length in whole seconds. |
| resolution | No | string | 768p | 480p, 768p | Native generation resolution. |
| prompt_expansion_mode * | Yes (*) | string | balanced | balanced, quality | How much effort to spend rewriting the prompt before generation. |
| seed | No | integer | random | Integer | Fixed seed for reproducible results; omit for a random seed. |
| enable_safety_checker | No | boolean | true | true, false | Enables the safety checker when true. |
| Resolution | Price |
|---|---|
| 480p | $0.055 per second |
| 768p | $0.088 per second |
Billing for MiniMax H3 Max follows the generated clip duration. Check the Generation section on this page for the live credit estimate.
480p for cheaper MiniMax H3 Max drafts or 768p for sharper delivery.balanced or quality prompt expansion, then generate.quality expansion when the brief is sparse; use balanced when the prompt is already detailed.When you need pure text-to-video instead of a still, open the sibling MiniMax H3 Max text page on RunComfy. This image-to-video route is the place to keep an approved first frame while MiniMax H3 Max handles motion and sound.
Studios often keep MiniMax H3 Max in the loop for both concept boards and still-driven product shots so the look stays consistent across modes.
For product updates beyond this MiniMax H3 Max page:
Turn photos into expressive videos with synced voice motion.
Generate cinematic 4K clips from prompts with audio sync and pro control
Image-to-video 3-15s clips at $0.083 per second.
Seedance 2.5 Reference to Video: Turn reference images, videos, and audio into cinematic AI video
Enhanced 1080p image motion conversion for expressive, fluid video creation
LTX 2.5 Pro: still-to-video with synced audio, quality mode
MiniMax H3 Max is built for short clips where picture and sound are created together from a text brief or a first-frame still. It is a strong fit when you need clear camera moves, ordered action beats, and matching audio without a separate sound pass.
Yes. MiniMax H3 Max covers text-to-video when you invent the first frame from words, and image-to-video when you lock an opening still. This page focuses on image-to-video; use the sibling text-to-video page for the text-only path.
On the image-to-video route, MiniMax H3 Max accepts an optional end image so the clip travels from your start frame toward a chosen closing frame. Keep subject identity, wardrobe, and palette consistent between the two stills for cleaner transitions.
MiniMax H3 Max generates clips at 480p or 768p with duration from 5 to 15 seconds. Choose 480p for cheaper drafts and 768p when you need sharper delivery; check the RunComfy parameter panel for the live controls.
MiniMax H3 Max offers balanced and quality prompt-expansion modes. Balanced rewrites the brief quickly; quality spends longer building a richer prompt before generation. Use quality for sparse briefs and balanced when your prompt is already detailed.
Yes. MiniMax H3 Max returns synchronized audio with the picture, so you can steer ambience, foley, or dialogue cues in the same prompt. If you omit audio direction, the model still supplies a soundtrack based on the scene.
Yes. Prototype MiniMax H3 Max in the RunComfy Web UI, then call the same model through the RunComfy API with identical parameters for production or automation. You do not need to host or scale the model yourself.
MiniMax H3 Max is billed per second of generated video: $0.055 per second at 480p and $0.088 per second at 768p. Generations consume USD or credits; check the Generation section on this page for the live estimate before you run.
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.
