Hailuo 2.3 Physics Realism

Hailuo 2.3 is a video generation model from MiniMax for text-to-video and image-to-video workflows. It improves character motion, physical actions, visual quality, stylization, and facial expressions over Hailuo 02. Atlas Cloud provides access through dedicated Standard, Pro, and Fast model endpoints.

Turn a written prompt or a still image into a short video with controlled movement and visual styling. The model is particularly suited to character action, expressive faces, anime, illustration, ink wash, game CG, and cinematic effects. Image-to-video generation uses the supplied image as the first frame.

Create an API key in the Atlas Cloud dashboard and send an authenticated POST request to `https://api.atlascloud.ai/api/v1/model/generateVideo`. Include the selected model ID and either a prompt or an image, depending on the endpoint. The request starts an asynchronous generation job and returns an ID for tracking.

Atlas Cloud offers `t2v-standard` and `t2v-pro` for text-to-video, plus `i2v-standard` and `i2v-pro` for image-to-video. The separate `minimax/hailuo-2.3/fast` model prioritizes faster generation for rapid iteration. Select the endpoint by input type first, then balance speed, fidelity, and price.

For text-to-video, provide the model ID and a prompt of up to 2,000 characters, with duration and prompt expansion available as optional controls. Image-to-video requires a JPG, JPEG, or PNG image supplied through a public URL or Base64 data URI, while the prompt is optional. Images must be under 20 MB, have a short side above 300 pixels, and use an aspect ratio between 2:5 and 5:2.

MiniMax documents the model family with 768p and 1080p output and durations of 6 or 10 seconds. Atlas Cloud exposes 6-second and 10-second duration values on its Standard text-to-video and image-to-video endpoints. Available resolution and duration combinations can differ by variant, so check the selected endpoint schema before submitting a production job.

Video generation begins when the API accepts your request and returns a prediction ID with its current status. Poll the Atlas Cloud prediction endpoint with that ID until the job reaches a terminal state. Once processing succeeds, read the generated video location from the response output.

Atlas Cloud lists a standard base price of $0.19 for Fast, $0.28 for Standard text-to-video or image-to-video, and $0.49 for Pro text-to-video or image-to-video. These are pay-as-you-go base prices rather than subscription commitments. Choose Fast for inexpensive iterations, then move to Standard or Pro when the job needs greater production quality.

Choose Fast when turnaround and iteration cost matter most. Standard balances visual quality, prompt fidelity, and price for routine creative workflows, while Pro targets detailed scene evolution, stronger continuity, advanced motion, physics realism, and film-style output. Test the same brief across tiers before committing a large batch.

No, the current model produces video without native audio. Add dialogue, music, or sound effects afterward with an editing workflow or a separate audio model. This also lets you replace or localize the soundtrack without regenerating the visuals.