Last verified: August 13, 2026.
Quick answer: MiniMax Hailuo 2.3 is now listed by MiniMax as a legacy V1 video model. Its exact API ID remains MiniMax-Hailuo-2.3, with documented text-to-video and image-to-video support: 6 seconds at 768P or 1080P, or 10 seconds at 768P, at 24 fps. For a new integration, evaluate the current MiniMax-H3 model on the V2 video API first.
Independent-site notice: MiniMax-AI.chat is an independent educational website. It is not MiniMax’s official website and is not endorsed by or affiliated with MiniMax. Product names and trademarks belong to their respective owners. API access, interface availability, billing, and provider terms are controlled by MiniMax.
On this page
- What Hailuo 2.3 is
- Verified specifications
- Text-to-video
- Image-to-video
- Prompting and camera commands
- API workflow and Node.js example
- API pricing
- Limitations and responsible use
- Frequently asked questions
What is MiniMax Hailuo 2.3?
MiniMax Hailuo 2.3 is a generative video model from MiniMax. The company announced it on October 28, 2025 as a successor built on Hailuo 02. The official Hailuo 2.3 release article describes vendor-reported improvements in physical action, character expression, stylized output, and response to motion instructions. Those statements describe the provider’s release claims; they are not independent benchmark results.
The current official MiniMax model catalog places Hailuo 2.3 under Legacy Models. Its V1 references still document two input modes:
- Text-to-video: start from a required written prompt.
- Image-to-video: start from a required first-frame image and optionally describe the motion with a prompt.
This API capability does not prove that a particular consumer website, mobile app, region, account tier, or third-party tool exposes the same model selector. Check the interface you plan to use before purchasing credits or designing a production workflow.
MiniMax Hailuo 2.3 specifications
| Specification | Verified value |
|---|---|
| Official API model ID | MiniMax-Hailuo-2.3 |
| Documented input modes | Text-to-video and image-to-video |
| Frame rate | 24 fps |
| 768P duration | 6 or 10 seconds |
| 1080P duration | 6 seconds |
| Default duration | 6 seconds |
| Default resolution for this model | 768P |
| Prompt limit | Up to 2,000 characters |
| Task type | Asynchronous |
| Generation endpoint | POST https://api.minimax.io/v1/video_generation |
The labels 768P and 1080P are the documented API resolution values. Do not translate them into a fixed width-by-height promise without reading the returned video_width and video_height for the completed task.
Standard model versus Hailuo 2.3 Fast
| Question | Hailuo 2.3 | Hailuo 2.3 Fast |
|---|---|---|
| Exact API ID | MiniMax-Hailuo-2.3 | MiniMax-Hailuo-2.3-Fast |
| Text-to-video listed in official API docs? | Yes | No |
| Image-to-video listed? | Yes | Yes |
| 768P durations | 6 or 10 seconds | 6 or 10 seconds |
| 1080P duration | 6 seconds | 6 seconds |
| PAYG price: 768P, 6 seconds | $0.28 | $0.19 |
Within legacy V1 integrations, choose the standard Hailuo 2.3 ID when text-to-video is required; both Hailuo 2.3 IDs remain documented for image-to-video. For a new video integration, evaluate current H3/V2 first. Price and input support are objective differences; the best visual result still depends on the source image, motion brief, subject, and review criteria.
Text-to-video with Hailuo 2.3
The official text-to-video API reference lists MiniMax-Hailuo-2.3 as a supported model. A text-to-video request requires model and prompt. The prompt can contain no more than 2,000 characters.
| Field | Requirement for Hailuo 2.3 |
|---|---|
model | Required; use the exact case-sensitive value MiniMax-Hailuo-2.3 |
prompt | Required; maximum 2,000 characters |
duration | Optional; defaults to 6; use 6 or 10 at 768P, and 6 at 1080P |
resolution | Optional; defaults to 768P; 1080P is available for 6 seconds |
prompt_optimizer | Optional boolean; defaults to true |
fast_pretreatment | Optional boolean; defaults to false; reduces optimization time when prompt optimization is enabled |
callback_url | Optional endpoint URL for asynchronous status updates |
fast_pretreatment is a request option, not the Fast model ID. Setting it to true does not change MiniMax-Hailuo-2.3 into MiniMax-Hailuo-2.3-Fast.
Image-to-video with Hailuo 2.3
The official image-to-video API reference supports the same standard model ID. Here, first_frame_image is required and prompt is optional.
| Image rule | Documented requirement |
|---|---|
| How to provide the image | A publicly reachable URL or a Base64 data URL such as data:image/jpeg;base64,... |
| Accepted formats | JPG, JPEG, PNG, or WebP |
| Maximum file size | Less than 20 MB |
| Minimum dimensions | Short edge must be greater than 300 pixels |
| Allowed source-image aspect ratio | Between 2:5 and 5:2 |
| Prompt | Optional; maximum 2,000 characters |
The image is the opening frame. The ratio requirement above is an input-validation rule; it should not be rewritten as a universal output-dimension guarantee. Read video_width and video_height from the successful status response.
Prepare a better first frame
- Use a sharp, well-exposed image with a clearly readable subject.
- Leave plausible room for the movement you request. A subject pressed against the edge has less visual space to move.
- Keep anatomy, small objects, and text legible before animation; generation may amplify flaws in the source.
- Request one coherent action and one camera plan for a short clip.
- Use only images you are authorized to upload and animate.
How to prompt MiniMax Hailuo 2.3
A useful short-video prompt describes what can be seen and how it changes. A practical order is: subject + action + environment + camera movement + lighting + visual treatment + ending state. Avoid stuffing several scene changes into six seconds.
Text-to-video prompt examples
- Product shot: “A matte ceramic perfume bottle stands on pale limestone. A narrow band of morning light moves across the label while fine dust drifts in the air. [Push in] Clean studio composition, realistic reflections, the bottle remains centered.”
- Landscape: “A small red canoe crosses a misty alpine lake at dawn. Gentle ripples expand behind it while low clouds reveal dark pine trees. [Pan right] Natural color, restrained motion, wide cinematic framing.”
- Stylized scene: “A paper-cut fox walks through layered blue forest shapes as leaves sway in a light breeze. [Tracking shot] Handcrafted paper texture, limited color palette, soft side lighting.”
Image-to-video motion prompts
- “The subject takes one slow step forward, fabric moving gently in the breeze. [Pull out] Preserve facial features, clothing colors, and the background layout.”
- “Steam rises from the cup and curls toward the window light. The camera remains locked. [Static shot] Keep the cup shape and printed mark stable.”
- “The illustrated clouds drift from left to right while the foreground flowers bend slightly. [Pedestal up] Preserve the original ink-and-watercolor style.”
Documented camera commands
Hailuo 2.3 accepts 15 explicit commands in square brackets: [Truck left], [Truck right], [Pan left], [Pan right], [Push in], [Pull out], [Pedestal up], [Pedestal down], [Tilt up], [Tilt down], [Zoom in], [Zoom out], [Shake], [Tracking shot], and [Static shot].
Commands placed together in one bracket are simultaneous; MiniMax recommends no more than three in that bracket. Commands placed at different points in the prompt are sequential. Natural-language camera descriptions also work, while the reference says explicit commands provide more accurate control. Treat a command as an instruction, not a guarantee that every frame will follow it perfectly.
MiniMax Hailuo 2.3 API workflow
Video generation is asynchronous. The official video-generation workflow uses three stages:
- Send a generation request and store the returned
task_id. - Query
GET /v1/query/video_generation?task_id=...until the status isSuccessorFail. The guide recommends a 10-second polling interval. - On success, use the returned
file_idwithGET /v1/files/retrieve?file_id=..., then readfile.download_url.
Polling responses use the case-sensitive states Preparing, Queueing, Processing, Success, and Fail. Callback payloads use lowercase processing, success, and failed. If you configure callback_url, your server must echo the validation challenge within three seconds.
For a broader explanation of authentication, errors, and production controls, see our independent MiniMax API guide.
MiniMax Hailuo 2.3 API pricing
The official MiniMax pay-as-you-go pricing page showed these legacy-model rates when verified on August 13, 2026:
| Model | Resolution | Duration | Price per generated video |
|---|---|---|---|
MiniMax-Hailuo-2.3 | 768P | 6 seconds | $0.28 |
MiniMax-Hailuo-2.3 | 768P | 10 seconds | $0.56 |
MiniMax-Hailuo-2.3 | 1080P | 6 seconds | $0.49 |
These are official API PAYG prices, not a promise about subscriptions, promotional credits, taxes, currency conversion, consumer Hailuo plans, or third-party markups. Confirm the provider’s billing page before running a large batch. Our MiniMax pricing explainer separates API rates from other access routes.
Limitations, review checks, and responsible use
Documented capability boundaries
- No 10-second 1080P mode: 10 seconds is documented only at 768P.
- No 512P option for Hailuo 2.3: the model’s documented choices are 768P and 1080P.
- No documented first-and-last-frame support: the official first-and-last-frame API reference lists only
MiniMax-Hailuo-02, not Hailuo 2.3. - No audio promise: the video-generation references covered here do not describe synchronized speech, music, or sound generation as part of a Hailuo 2.3 request.
- Asynchronous delivery: request acceptance is not video completion. Your integration must handle queues, failures, timeouts, and retrieval.
- Interface availability varies: API documentation does not guarantee access in every MiniMax interface, plan, country, or reseller.
Review every generated clip
Generative video can produce identity drift, unstable hands, changing object counts, warped logos, unreadable small text, flicker, implausible motion, or unintended background details. Check the full clip frame by frame before publication. For brand work, compare colors, packaging, marks, and factual product features against approved references. A good-looking preview is not proof of factual, legal, or safety suitability.
Rights, consent, disclosure, and privacy
- Upload only images and prompts you own, license, or have permission to use.
- Obtain appropriate consent before animating an identifiable person. Apply heightened safeguards to images of children and vulnerable people.
- Do not use generated video for impersonation, fraud, harassment, sexual exploitation, or deceptive evidence.
- Check copyright, trademark, likeness/publicity, music, and advertising rules for the intended territory and platform.
- Disclose synthetic or AI-generated media when a law, platform policy, contract, or context requires it.
- Do not place confidential, regulated, or private images in an external service unless your organization has approved the data flow and controls. See our sensitive-data checklist.
The API references reviewed for this page do not establish a universal visible-watermark rule for every delivered Hailuo 2.3 file. Inspect the actual output and follow the provider’s applicable terms and the destination platform’s labeling requirements. This page does not guarantee ownership, exclusivity, commercial clearance, or legal compliance for any output.
MiniMax Hailuo 2.3 FAQ
What is the exact MiniMax Hailuo 2.3 API model ID?
Use MiniMax-Hailuo-2.3. Preserve capitalization and punctuation exactly. Do not substitute a marketplace slug or an unofficial provider alias when calling MiniMax’s own endpoint.
Does Hailuo 2.3 support text-to-video?
Yes. The official MiniMax text-to-video reference lists the standard MiniMax-Hailuo-2.3 model and requires a prompt. The Fast model is not listed on that route.
Does Hailuo 2.3 support image-to-video?
Yes. Supply a valid first_frame_image by public URL or Base64 data URL. The prompt is optional and can contain up to 2,000 characters.
Can Hailuo 2.3 make a 10-second 1080P video?
No documented combination supports that. Use 10 seconds at 768P, or 6 seconds at 768P or 1080P.
Can I specify both a first and last frame?
Not with Hailuo 2.3 according to the official first-and-last-frame reference verified for this page. That route lists MiniMax-Hailuo-02 only.
Does prompt optimization change the model?
No. prompt_optimizer controls automatic prompt optimization, and fast_pretreatment can reduce that optimization stage when enabled. Neither field changes the requested model ID.
Conclusion: Hailuo 2.3 remains a documented legacy V1 option for existing integrations that need both text-to-video and image-to-video. New builds should evaluate MiniMax-H3 on V2 first; do not reuse Hailuo’s V1 payload, polling, or pricing assumptions for H3. If maintaining Hailuo 2.3, keep its exact short-clip combinations, asynchronous task lifecycle, and per-generation cost, then apply human review, rights checks, and disclosure rules before publishing.
More on Hailuo: see the Hailuo AI hub for every model compared side by side, or Is Hailuo AI free? for what each access route actually costs.