Hailuo 2.3: Complete Specifications, Pricing, API Access & Use Cases (2026)

What Is Hailuo 2.3?

Hailuo 2.3 is MiniMax’s short-form video generation model, released on October 28, 2025, featuring text-to-video and image-to-video generation, 6–10 second output options, and 768p or 1080p resolution choices, with Gate.AI listing pricing by duration and resolution as of July 2026. MiniMax introduced Hailuo 2.3 as an update to Hailuo 02 and described improvements in dynamic expression, physical actions, stylization, character micro-expressions, and response to motion commands.

Hailuo 2.3 is not a large language model for chat, reasoning, or coding. It is a generative video model for producing short video clips from text prompts or source images. This makes it relevant for creative teams, social media producers, e-commerce marketers, storyboard artists, and developers building short-form video generation workflows.

MiniMax’s official video API documentation describes video generation as an asynchronous process: create a task, poll the task status, and retrieve the generated file after completion. Gate.AI’s public documentation also describes an asynchronous video generation endpoint with task submission, polling, billing fields, and download URL retrieval.

What Are Hailuo 2.3’s Key Specifications and Pricing?

The following table separates provider-verified information from Gate.AI listing details. MiniMax’s official pricing documentation uses video points, while the Gate.AI listing for this model presents dollar pricing by resolution and duration. MiniMax states that MiniMax-Hailuo-2.3 deducts 1 video point for a 768p 6-second video, 2 video points for a 768p 10-second video, and 2 video points for a 1080p 6-second video.

What Can Hailuo 2.3 Do That Makes It Useful in Production?

Hailuo 2.3 can generate short videos from text prompts, making it useful for fast visual ideation, campaign mockups, storyboard tests, and short-form creative drafts. MiniMax’s video generation workflow supports asynchronous generation, which fits production systems that submit tasks, poll status, and retrieve completed media files.

The model also supports image-to-video generation. MiniMax’s image-to-video API documentation shows a workflow where a prompt, first-frame image, model ID, duration, and resolution are used to create a video task. This is useful when a team already has an approved product image, character frame, concept still, or brand visual and wants to add controlled motion.

MiniMax describes Hailuo 2.3 as improving dynamic body movement, motion command following, facial micro-expression rendering, stylization, lighting behavior, and object motion response. These capabilities matter in short video production because motion consistency, subject stability, and camera-control adherence often determine whether generated clips are usable without excessive reruns.

Hailuo 2.3 may be especially relevant for AI short-video generation workflows where teams need compact clips, predictable pricing, and prompt-controlled camera movement. Related model profiles such as Seedance 2.0 video generation can help to compare different video model families.

What Are Hailuo 2.3’s Supported Modalities?

Where Does Hailuo 2.3 Fall Short?

Hailuo 2.3 is designed for short-form video, not long-form scene continuity. The Gate.AI listing covers 6-second and 10-second generation options, while MiniMax’s official pricing documentation lists point deductions for 768p/6s, 768p/10s, and 1080p/6s. For 1080p/10s, the Gate.AI listing does not provide a price, so teams should verify availability before building fixed-cost workflows around that configuration.

Pricing also requires careful source separation. MiniMax official documentation uses video points, while the Gate.AI listing uses dollar prices for specific duration-resolution combinations. These are separate billing presentations and should not be merged unless a pricing page explicitly maps one unit to the other.

This is a general AI limitation and is not model-specific unless stated by the provider: generated videos may contain artifacts, visual distortions, unrealistic physics, identity drift, prompt misunderstanding, or unexpected object changes. Human review is necessary before using generated content in brand, legal, political, medical, financial, safety-sensitive, or rights-sensitive environments.

Copyright and likeness review are also important for AI video generation. Production teams should avoid prompts that request protected characters, real-person likenesses without permission, logos, copyrighted scenes, or other restricted assets unless they have the necessary rights and policy clearance.

What Is Hailuo 2.3 Best Used For?

How Does Hailuo 2.3 Compare to Seedance 2.0 and Veo 3.1?

Seedance 2.0 and Veo 3.1 are useful comparison models because they target modern AI video generation with stronger multimodal and audio-video workflows. ByteDance states that Seedance 2.0 supports text, image, audio, and video input modalities and up to 15-second high-quality multi-shot audio-video output. Google describes Veo 3.1 as supporting text-to-video, image-to-video, text-to-audio-plus-video generation, and realistic physics, with Gemini API access for 8-second videos with native audio.

How Do I Access Hailuo 2.3 Through Gate.AI?

As per Gate.AI listing, Hailuo 2.3 is available with the model ID minimax/hailuo-2.3. Gate.AI’s public documentation verifies the asynchronous video task API at POST /api/v1/videos, Bearer-token authentication, optional idempotency keys for safe retries, video prompt fields, duration, resolution, aspect ratio, reference media fields, webhook support, and task polling through GET /api/v1/videos/{job_id}.

Gate.AI’s pricing page states that video and other multimodal capabilities are billed by generation count, duration, resolution, or task specification. For Hailuo 2.3 specifically, the Gate.AI listing shows 768p/6s at \$0.28, 768p/10s at \$0.56, and 1080p/6s at \$0.49 as of July 2026.

Python Example

import os
import uuid
import requests
api_key = os.environ["GATE_AI_API_KEY"]
headers = {
    "Authorization": f"Bearer {api_key}",
    "Content-Type": "application/json",
    "Idempotency-Key": str(uuid.uuid4()),
}
payload = {
    "model": "minimax/hailuo-2.3",
    "prompt": (
        "A cinematic product shot of a matte black electric scooter "
        "turning slowly under neon city lights, smooth camera pan, "
        "stable subject, realistic reflections."
    ),
    "duration": 6,
    "resolution": "1080p",
    "aspect_ratio": "16:9",
    "generate_audio": False
}
response = requests.post(
    "https://api.gate.ai/api/v1/videos",
    headers=headers,
    json=payload,
    timeout=60
)
response.raise_for_status()
job = response.json()["data"]
print("Job ID:", job["job_id"])
print("Status URL:", job["status_url"])

curl Example

curl https://api.gate.ai/api/v1/videos \
  -H "Authorization: Bearer $GATE_AI_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "model": "minimax/hailuo-2.3",
    "prompt": "A cinematic product shot of a matte black electric scooter turning slowly under neon city lights, smooth camera pan, stable subject, realistic reflections.",
    "duration": 6,
    "resolution": "1080p",
    "aspect_ratio": "16:9",
    "generate_audio": false
  }'

To check generation progress, use the returned job_id with GET /api/v1/videos/{job_id}. Gate.AI’s documented response includes status, model, status URL, download URL after completion, duration, resolution, estimated cost, billed cost, billing status, and timestamps.

FAQs

What is Hailuo 2.3?
Hailuo 2.3 is MiniMax’s short-form AI video generation model for text-to-video and image-to-video workflows. It is designed for short creative clips, motion-controlled prompts, subject stability, stylized visuals, and image-based video animation.

How much does Hailuo 2.3 cost on Gate.AI?
As per Gate.AI listing, Hailuo 2.3 costs \$0.28 for 768p/6s, \$0.56 for 768p/10s, and \$0.49 for 1080p/6s as of July 2026. A 1080p/10s price is not listed.

Can developers access Hailuo 2.3 through an API?
Yes. MiniMax documents video generation through its API, and Gate.AI documents an asynchronous video generation endpoint. As per Gate.AI listing, the model ID is minimax/hailuo-2.3.

When should teams choose Hailuo 2.3 instead of Seedance 2.0 or Veo 3.1?
Hailuo 2.3 may fit short text-to-video and image-to-video workflows where duration, resolution, and iteration cost are central. Seedance 2.0 or Veo 3.1 may fit projects that require stronger audio-video or multimodal reference workflows.