Develop a reference-led shot
Use several assets to explain the scene. Identify which reference defines the product and which sets the visual direction.
MiniMax Video generation
Create video with MiniMax H3 in Riverflow, using a written brief, start and end frames, or multiple visual references.
Open RiverflowThe example
A finished product image, a camera direction and the generated result. Play the clip, then open the prompt to see how it was made.
A slow, precise camera orbit moves from the front three-quarter view towards the side of this orange suede running shoe. The shoe stays perfectly still on the limestone. Gentle blue sea ripples move in the background and a natural glint travels across the amber sole. Warm late-afternoon light, realistic texture, subtle shallow depth of field. One continuous product shot, no cuts, no added objects, no lettering or shape changes. Preserve product identity, shape, and package proportions. Keep brand text and logos readable; avoid hallucinated text or misspellings. Use smooth natural motion and coherent camera movement with no warping. Maintain commercial-grade lighting consistency and realistic material behavior. Do not alter core packaging design elements or create extra products.
Select MiniMax H3 after opening the video tool.
MiniMax H3 is a multimodal video generation model from MiniMax. It interprets text, images, video and audio together to create footage with native stereo sound, with a focus on reference control and instruction following. Riverflow supports H3 at 2K, including text-to-video, frame-guided generation and multi-reference generation.
Choose MiniMax H3 in Riverflow’s video tool, add your brief and references, then review the settings and credit cost before generating.
Available settings depend on the model and workflow.
Plans and credits| Model | 5s | 10s |
|---|---|---|
| MiniMax H32K · 16:9 · starting image · audio off | 45 credits | 90 credits |
These are generation costs for a starting image that already fits the output. Preparing a mismatched frame can add 20 credits per image. Resolution, aspect ratio, audio and reference inputs can change the total. Your complete quote appears before generation.
Use several assets to explain the scene. Identify which reference defines the product and which sets the visual direction.
Start from a composed product image. A slow camera movement can reveal its surfaces, depth and silhouette.
Supply two compatible frames to guide the transition. Review the movement between them, as well as the endpoints.
Models, inputs and getting started.
MiniMax H3 is a multimodal video generation model from MiniMax. It interprets text, images, video and audio together to create footage with native stereo sound, with a focus on reference control and instruction following. Riverflow supports H3 at 2K, including text-to-video, frame-guided generation and multi-reference generation.
Yes. Riverflow supports a multi-reference mode for MiniMax H3, alongside text and frame-guided modes. Follow the input requirements shown by the chosen workflow; do not assume it shares Seedance's reference limits or settings.
Sign in to Riverflow, open the video creation tool and choose MiniMax H3 from the model picker. Add your brief and any required references, then review the available settings and credit cost before generating.
MiniMax H3: 45 credits for 5s, 90 credits for 10s at 2k · 16:9 · starting image · audio off. These are generation costs; reference preparation, resolution, aspect ratio, audio and video-input workflows can change the total. Riverflow shows the full quote before you generate.
Bring your brief. Choose your model. Start creating.