Last verified: August 13, 2026.
Quick answer: MiniMax Hailuo 2.3 Fast is now listed by MiniMax as a legacy V1 image-to-video model. Its exact API ID remains MiniMax-Hailuo-2.3-Fast; it requires a first-frame image and supports 6 seconds at 768P or 1080P, or 10 seconds at 768P, at 24 fps. Within the legacy Hailuo family, use standard Hailuo 2.3 when text-only generation is required. For a new integration, evaluate current MiniMax-H3 on V2 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. MiniMax controls API access, billing, availability, and provider terms.
On this page
- What Hailuo 2.3 Fast is
- Verified specifications
- Image requirements
- Standard versus Fast
- Prompting and camera control
- API workflow and Node.js example
- API pricing
- Limitations and responsible use
- Frequently asked questions
What is MiniMax Hailuo 2.3 Fast?
MiniMax Hailuo 2.3 Fast is the image-to-video member of the Hailuo 2.3 family. It turns a supplied image into the opening frame of a short generated clip. An optional text prompt describes the intended subject motion, environmental motion, camera movement, and visual treatment.
The official Hailuo 2.3 announcement, published October 28, 2025, presents Fast as a faster, lower-priced variant. “Faster” is a provider positioning statement, not a fixed latency guarantee: queue load, account limits, region, input, and request settings can affect completion time. The official PAYG table does independently establish lower per-video prices for Fast at each matching Hailuo 2.3 resolution-duration combination.
The current MiniMax model catalog places Hailuo 2.3 Fast under Legacy Models and lists it for image-to-video at 24 fps, not text-to-video. MiniMax lists MiniMax-H3 as the current video model on V2. Fast therefore remains a distinct legacy V1 model ID, not a speed switch that can be attached to any Hailuo or H3 request.
API documentation also does not prove that the model appears in a particular Hailuo consumer interface, mobile app, country, subscription, or third-party platform. Verify the actual product surface and account before committing budget or promising access to users.
MiniMax Hailuo 2.3 Fast specifications
| Specification | Verified value |
|---|---|
| Official API model ID | MiniMax-Hailuo-2.3-Fast |
| Documented input mode | Image-to-video |
| Required visual input | One first_frame_image |
| Prompt | Optional; maximum 2,000 characters |
| Frame rate | 24 fps |
| 768P duration | 6 or 10 seconds |
| 1080P duration | 6 seconds |
| Default duration | 6 seconds |
| Default resolution | 768P |
| Task type | Asynchronous |
| Generation endpoint | POST https://api.minimax.io/v1/video_generation |
Resolution labels do not by themselves promise one fixed width and height across every input. Read video_width and video_height from the successful task response rather than inventing pixel dimensions.
First-frame image requirements
The official MiniMax image-to-video API reference requires first_frame_image for the Fast model.
| Rule | Documented requirement |
|---|---|
| Input method | Publicly reachable image URL or Base64 data URL |
| Formats | JPG, JPEG, PNG, or WebP |
| File size | Less than 20 MB |
| Short edge | Greater than 300 pixels |
| Aspect ratio | Between 2:5 and 5:2 |
| Role in the video | Opening frame |
The 2:5-to-5:2 rule validates the source image. It should not be turned into a claim that every output dimension is automatically identical to the source. Inspect the dimensions reported after generation.
Choose a first frame that can move
- Start with a sharp, correctly oriented image and one visually dominant subject.
- Leave space in the direction of travel, a camera pan, or a pull-back.
- Avoid tiny lettering and fine logos when exact legibility is essential; generated frames can distort them.
- Remove accidental overlays, malformed hands, and background defects before submitting the image.
- Keep the requested action plausible for the subject’s pose and the available six or ten seconds.
- Confirm that you are authorized to upload and animate the image.
MiniMax Hailuo 2.3 versus Hailuo 2.3 Fast
| Decision point | Hailuo 2.3 | Hailuo 2.3 Fast |
|---|---|---|
| Exact API ID | MiniMax-Hailuo-2.3 | MiniMax-Hailuo-2.3-Fast |
| Text-to-video in official docs | Supported | Not listed |
| Image-to-video in official docs | Supported | Supported |
| 768P, 6 seconds | $0.28 | $0.19 |
| 768P, 10 seconds | $0.56 | $0.32 |
| 1080P, 6 seconds | $0.49 | $0.33 |
| Prompt limit for image-to-video | 2,000 characters | 2,000 characters |
Within a legacy Hailuo V1 integration, use the Fast ID when the input is an image and its documented price or provider-positioned speed fits the workflow; use the standard Hailuo 2.3 ID when text-only generation is needed. For a new integration, evaluate H3/V2 first. Do not claim that one model always has better visual quality without a controlled test on your own representative inputs; the API references do not publish a universal quality ranking between them.
Fast model ID versus fast pretreatment
These are separate concepts:
MiniMax-Hailuo-2.3-Fastselects the Fast model.fast_pretreatmentis an optional boolean request field that reduces optimization time whenprompt_optimizeris enabled. It defaults tofalseand is documented for both Hailuo 2.3 variants.
Turning on fast_pretreatment does not rewrite the model ID. Set the intended model explicitly in every request.
How to prompt Hailuo 2.3 Fast
The image already defines the starting composition, subject appearance, and visible environment. Use the optional prompt primarily to specify subject motion + environmental motion + camera plan + elements to preserve + ending state. Short, compatible instructions are easier to evaluate than a chain of unrelated actions.
Image-to-video prompt examples
- Portrait: “The subject turns slightly toward the window and blinks once; loose hair moves in a gentle breeze. [Push in] Preserve facial features, clothing, and background geometry.”
- Food: “Steam rises naturally from the bowl while the herbs shift slightly. [Static shot] Preserve the dish shape, table setting, and label placement.”
- Product: “A soft highlight travels from left to right across the metal surface. The object rotates a few degrees clockwise. [Pull out] Keep proportions, engraving, and brand colors stable.”
- Illustration: “The paper birds flap slowly and cross behind the central tree as the foreground grass sways. [Pan right] Preserve the flat cut-paper texture and limited palette.”
- Architecture: “Cloud shadows move across the facade and the trees sway lightly. [Pedestal up] Straight vertical lines, realistic daylight, no change to windows or signage.”
Supported camera commands
The image-to-video reference documents 15 commands for Hailuo 2.3 Fast: [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].
Multiple commands inside one bracket happen simultaneously; the provider recommends a maximum of three there. Commands placed in order throughout the prompt are sequential. Natural-language directions also work, but the reference describes bracketed commands as more accurate. No command eliminates the need to inspect the result.
Hailuo 2.3 Fast API workflow
The request uses POST https://api.minimax.io/v1/video_generation. MiniMax returns a task_id, not the finished video. The official asynchronous video guide then polls the task and retrieves the completed file:
- Create the image-to-video task with the Fast model ID and first-frame image.
- Poll
GET /v1/query/video_generation?task_id=...at the recommended 10-second interval. - Continue through
Preparing,Queueing, orProcessing. - On
Success, takefile_idand requestGET /v1/files/retrieve?file_id=.... - Read
file.download_url. OnFail, log the error and stop rather than resubmitting indefinitely.
If you use callback_url, MiniMax first sends a challenge that your endpoint must echo within three seconds. Callback status values are lowercase—processing, success, and failed—while the polling states above use title case. Keep these parsers separate. See our independent API implementation guide for additional authentication and error-handling context.
MiniMax Hailuo 2.3 Fast API pricing
The official MiniMax PAYG pricing page listed these legacy Fast rates when verified on August 13, 2026:
| Model | Resolution | Duration | Price per generated video |
|---|---|---|---|
MiniMax-Hailuo-2.3-Fast | 768P | 6 seconds | $0.19 |
MiniMax-Hailuo-2.3-Fast | 768P | 10 seconds | $0.32 |
MiniMax-Hailuo-2.3-Fast | 1080P | 6 seconds | $0.33 |
These amounts describe official API pay-as-you-go generation. They do not establish consumer subscription pricing, promotional credit value, taxes, foreign-exchange costs, or third-party platform fees. Recheck MiniMax billing before a batch and budget for rejected creative attempts as well as accepted outputs. Our MiniMax pricing explainer keeps API and non-API access routes separate.
Exact limitations and responsible use
What the Fast API does not document
- No text-only Fast route: the official text-to-video API reference lists standard Hailuo 2.3, but not
MiniMax-Hailuo-2.3-Fast. - No first-and-last-frame Fast route: the first-and-last-frame reference lists only
MiniMax-Hailuo-02. - No 512P Fast option: Fast is documented at 768P and 1080P.
- No 10-second 1080P combination: use 10 seconds at 768P, or 6 seconds at 768P or 1080P.
- No built-in audio promise: the Fast image-to-video request reference does not describe synchronized dialogue, music, or sound generation.
- No fixed completion-time promise: asynchronous queues and failures remain possible despite the Fast positioning.
- No cross-interface guarantee: API support does not ensure the same choice appears in a consumer UI or reseller.
Quality review before publication
Inspect every frame for facial drift, changed clothing or packaging, unstable hands, flicker, object duplication, warped architecture, invented text, distorted trademarks, and motion that contradicts the brief. Compare product and brand details against approved source material. Regenerate or edit when a clip could mislead viewers about a real person, place, event, or product.
Rights, consent, privacy, and disclosure
- Use an image only when you own it, license it, or have authorization to upload and animate it.
- Obtain appropriate consent before animating an identifiable person. Use heightened safeguards for children and vulnerable people.
- Do not create deceptive impersonations, fraudulent evidence, harassment, sexual exploitation, or illegal content.
- Review copyright, trademark, likeness/publicity, advertising, and destination-platform rules.
- Label or disclose synthetic media when required by law, contract, platform policy, or the risk of audience deception.
- Avoid sending confidential, regulated, or private images to an external provider without approved controls. Use our sensitive-data checklist before uploading.
The reviewed API references do not establish one universal visible-watermark rule for every Fast output. Inspect the delivered file and follow the provider’s terms and the publishing platform’s disclosure requirements. Generation does not guarantee ownership, exclusivity, commercial clearance, factual accuracy, or legal compliance.
MiniMax Hailuo 2.3 Fast FAQ
What is the exact Hailuo 2.3 Fast API model ID?
Use MiniMax-Hailuo-2.3-Fast with the exact capitalization and hyphenation shown. A third-party marketplace slug is not interchangeable with the official MiniMax API ID.
Does Hailuo 2.3 Fast support text-to-video?
Not according to the official MiniMax text-to-video reference verified for this page. Fast is listed for image-to-video. Use MiniMax-Hailuo-2.3 for the documented text-to-video route.
Is a prompt required?
No. The first-frame image is required, while the prompt is optional and can contain up to 2,000 characters. A concise motion prompt usually makes the creative intent easier to test and review.
Can Fast generate 1080P for 10 seconds?
No documented setting supports that combination. Fast supports 1080P for 6 seconds, or 768P for 6 or 10 seconds.
Can I give Fast both a start frame and an end frame?
Not through the official first-and-last-frame route verified here. That reference lists only MiniMax-Hailuo-02.
Is Fast always faster than standard Hailuo 2.3?
MiniMax positions Fast as faster, but the cited materials do not promise a fixed completion time or a guaranteed speed ratio. Measure end-to-end latency under your own account, request mix, concurrency, and region.
Is fast_pretreatment the same as the Fast model?
No. The model is selected with model: "MiniMax-Hailuo-2.3-Fast". fast_pretreatment is a separate prompt-optimization option and does not select or convert a model.
Conclusion: Hailuo 2.3 Fast remains a documented legacy V1 option for short image-to-video generation, with lower API prices than standard Hailuo 2.3 at matching settings. New builds should evaluate MiniMax-H3 on V2 first. If maintaining Fast, use its exact ID, validate the source image, and respect its documented resolution-duration pairs; do not extend its V1 support claims to text-to-video, first-and-last-frame, or H3.
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.