Cookie Preferences
We use cookies to personalize content and analyze traffic. See our Cookie Policy.
아직 동영상이 없습니다. 폼을 제출해 동영상을 생성하세요.
다양한 사용 사례와 매개변수 구성을 살펴보세요
숨겨진 수수료 없이 투명한 가격. 사용량에 따라 지불합니다.
Total Cost = Unit Price × (Generated Video Duration + Input Video Duration) + Additional Image Cost
| 규칙 및 양식 | Channel | 크레딧 | 가격 (USD) | 공식/참고 가격 | 일일 절감액 |
|---|---|---|---|---|---|
MiniMax-H3-video-input-768p videoMiniMaxresolution: 768P | Starter | 1per second | $0.005 | MiniMax$0.090per second | 일일 1,000 10초 비디오 기준MiniMax는 하루 $854.550 더 비쌉니다ApiPass가 94.95% 저렴 |
MiniMax-H3-text-to-video-2K videoMiniMaxresolution: 2K | Starter | 1per second | $0.005 | MiniMax$0.140per second | 일일 1,000 10초 비디오 기준MiniMax는 하루 $1354.550 더 비쌉니다ApiPass가 96.75% 저렴 |
MiniMax H3 사용 전체 가이드
Use MiniMax H3 API on ApiPass to turn prompts, frame images, and multimodal references into polished videos with up to 2K resolution, native stereo sound, consistent motion, and precise creative control.
MiniMax H3, also known as Hailuo 03, is a multimodal video generation system that understands text, images, video, and audio within a unified creative context. It supports text-to-video, image-to-video with first and last frame control, and reference-to-video using image, video, and audio assets. Through ApiPass, developers submit the public model name minimax/hailuo-h3, and ApiPass automatically selects the appropriate generation mode from the media fields in the input object.
Generate complete audiovisual clips directly from natural language prompts. Describe environments, characters, camera movement, shot progression, visual style, dialogue, music, and ambience in one request.
Animate a single first frame or create a controlled transition between first and last frames. This workflow is useful for product showcases, key-art animation, animated posters, UI demonstrations, and structured visual transitions.
Combine prompts with reference images, video clips, and audio files. Assign each asset to roles such as identity, product appearance, motion, camera language, editing rhythm, voice, or sound design.
MiniMax H3 can interpret text, images, video clips, and audio recordings within the same creative context, helping characters, motion, style, and sound remain connected throughout the generated video.
MiniMax H3 generates up to 2K video with native stereo sound, allowing visual action, dialogue, music, ambience, and sound effects to be directed together instead of added in a separate step.
Reference-to-video can combine multiple images, videos, and audio files in one request. Natural-language instructions clarify how each asset should influence identity, movement, style, voice, or atmosphere.
MiniMax H3 follows detailed directions for shot order, timing, transitions, camera movement, preservation requirements, readable text, and localized changes while maintaining unaffected scene details.
Create product films with close-up feature shots, branded typography, synchronized sound, and consistent materials across changing camera angles and scene transitions.
Generate opening sequences, animated posters, music visuals, editorial motion graphics, geometric layouts, typography-driven sequences, and sound-aware visual transitions.
Build short dramas, character promos, motion comics, and stylized narratives with consistent faces, clothing, props, voices, and performance details.
Prototype interface animations, game menus, equipment customization sequences, cinematic gameplay concepts, and digital experiences that combine interface logic with animated storytelling.
Choose the best MiniMax H3 workflow based on the source materials available for your application.
Use text-to-video when you want to start from a written prompt only. Provide prompt, duration, resolution, and a fixed aspect ratio such as 16:9 or 9:16.
Use image-to-video when you have a first frame or both first and last frames. ApiPass determines output ratio from the first frame, so aspect_ratio is not forwarded in this mode.
Use reference-to-video when you want images, videos, or audio to guide identity, motion, camera language, style, or sound. Reference mode supports adaptive and fixed aspect ratios.
Current regular pricing is 6 credits per second for 768P and 10 credits per second for 2K. A 5-second 2K generation costs 50 credits before any reference-media additions.
Reference video duration is added to billable seconds. Reference images after the first 5 add 6.4 credits each, while reference audio URLs currently do not add extra ApiPass H3 charges.
Failed tasks are refunded according to the existing task settlement logic. Use channel auto for automatic routing, or choose supported channels such as regular when you need more direct control.
Explore text-to-video, image-to-video, and reference-to-video in the ApiPass Playground. Test prompts, durations, resolutions, aspect ratios, first and last frames, and multimodal references before writing code.
Send a POST request to /api/v1/jobs/createTask using the public model name minimax/hailuo-h3. Provide prompt inside input, choose duration and resolution, and add frame or reference media when needed.
Use callBackUrl for success or failure notifications, and keep /api/v1/jobs/recordInfo as a polling fallback. Read generated video URLs from resultJson.resultUrls when the task state becomes success.
Use publicly accessible media URLs, validate mode conflicts before submission, keep prompts within 2000 characters for stable routing, and set reasonable polling intervals for active task retrieval.
모든 API는 Bearer Token을 통한 인증이 필요합니다.
Authorization: Bearer
Submit a MiniMax Hailuo H3 video generation task using text, frame images, or reference media.
The API accepts a JSON payload with the following structure:
1{
2 "model": "minimax/hailuo-h3",
3 "callBackUrl": "https://example.com/api/callback",
4 "channel": "auto",
5 "input": {
6 "prompt": "A cinematic ocean sunrise with realistic waves.",
7 "duration": 5,
8 "resolution": "2K",
9 "aspect_ratio": "16:9"
10 }
11}model필수stringPublic ApiPass model name. Must be minimax/hailuo-h3; do not pass minimax-h3, hailuo-03, or ApiPass model names with generation modes.
"minimax/hailuo-h3"
input필수objectH3 model input parameters. ApiPass uses these fields to normalize the request and automatically select text-to-video, image-to-video, or reference-to-video mode.
{"prompt":"A cinematic ocean sunrise.","duration":5,"resolution":"2K","aspect_ratio":"16:9"}callBackUrl선택적stringCallback URL for task success or failure notifications. The field name is case-sensitive and must be callBackUrl. If omitted, no callback will be sent.
"https://example.com/api/callback"
channel선택적stringProvider and billing channel used by APIPASS for the task. default channel is auto; APIPASS automatically allocates tasks across available providers based on real-time pricing and stability metrics; starter is ultra-low-cost with limited quotas and weaker stability; regular is standard and much cheaper than official APIs with moderate stability; official uses the model native API with high stability, fast task execution, and official pricing.
The default value for the channel parameter is auto. When enabled, APIPASS automatically allocates tasks across available providers based on real-time pricing and stability metrics to balance minimal cost and reliable performance. starter is an ultra-low-cost tier with limited quotas and weaker stability; regular is the standard tier and much cheaper than official APIs with moderate stability; official uses the model native API with high stability, fast task execution, and official pricing.
사용 가능한 옵션:
auto
fallback_models선택적arrayOptional ApiPass model-level fallback list. Each fallback item must provide its own model, channel, and complete input object.
[{"model":"minimax/hailuo-h3","channel":"auto","input":{"prompt":"A cinematic ocean sunrise.","duration":5,"resolution":"2K","aspect_ratio":"16:9"}}]The input object contains the following parameters:
input.prompt필수stringVideo generation prompt. It must be a non-empty string. ApiPass keeps up to 7000 characters; MInimax H3 API receives up to 7000 characters, while ApiPass receives only the first 2000 characters if the prompt is longer.
For cross-provider consistency, keep prompts within 2000 characters. In reference mode, refer to media by array order, such as Image 1, Video 1, or Audio 1.
"A cinematic ocean sunrise with realistic waves and smooth camera movement."
input.duration선택적integerOutput video duration in seconds. Public valid range is 4 to 15 inclusive, with a default of 5.
Invalid, missing, fractional, boolean, or out-of-range values are normalized to 5. ApiPass uses 5 seconds when the public request duration is 4.
사용 가능한 옵션:
5
input.resolution선택적stringOutput video resolution. Valid public values are 768P and 2K. Missing or invalid values default to 2K. The compatible alias quality maps to resolution, but resolution takes priority when both are present.
Values are trimmed and normalized to uppercase, so 768p and 2k are accepted. ApiPass currently outputs 2K even when the public request uses 768P.
사용 가능한 옵션:
"2K"
input.aspect_ratio선택적stringOutput video aspect ratio. Fixed values are 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. Reference-to-video also supports adaptive. Compatible aliases are aspectRatio and ratio, with priority aspect_ratio > aspectRatio > ratio.
Text-to-video defaults to 16:9 when missing or invalid. Image-to-video does not forward aspect_ratio because output ratio is determined by the first frame. Reference-to-video supports adaptive and fixed ratios.
사용 가능한 옵션:
"16:9"
input.first_frame_url선택적stringFirst-frame media URL for image-to-video mode. Compatible aliases by priority are first_frame_url, firstFrameUrl, image_url, imageUrl, and image.
Accepted as a string, an empty array, or a single-element array. Empty arrays and blank strings are treated as not provided; arrays with more than one element cause a parameter error.
"https://example.com/first.png"
input.last_frame_url선택적stringOptional last-frame media URL for image-to-video mode. Compatible alias: lastFrameUrl.
Use last_frame_url together with first_frame_url for consistent behavior across providers. It follows the same string and single-element array compatibility rules as first_frame_url.
"https://example.com/last.png"
input.reference_image_urls선택적array(string)Reference image URL array for reference-to-video mode. Maximum 9 elements. Empty strings are ignored during normalization. Compatible alias: referenceImageUrls.
Cannot be combined with first_frame_url, last_frame_url, or image_urls. Reference image URLs must be publicly accessible and should point to real media content.
["https://example.com/subject.png","https://example.com/environment.png"]
input.reference_video_urls선택적array(string)Reference video URL array for reference-to-video mode. Maximum 3 elements. ApiPass requires each reference video to be 2–15 seconds and total reference video duration to be no more than 15 seconds. Compatible alias: referenceVideoUrls.
The URL must be publicly accessible over http or https and no longer than 2048 characters. ApiPass reads actual reference video duration during task creation for accurate billing and rejects the task if duration cannot be read.
["https://example.com/motion.mp4"]
input.reference_audio_urls선택적array(string)Reference audio URL array for reference-to-video mode. Maximum 3 elements. Empty strings are ignored during normalization. ApiPass supports audio-only reference mode; ApiPass requires reference audio to be combined with reference images or videos. Compatible alias: referenceAudioUrls.
ApiPass does not support audio-only reference mode. If only reference audio is provided and the task routes to ApiPass, the adapter ignores the audio and falls back to text-to-video with a 16:9 ratio.
["https://example.com/atmosphere.mp3"]
1curl --location 'https://api.apipass.dev/api/v1/jobs/createTask' \
2 --header 'Authorization: Bearer YOUR_APIPASS_API_KEY' \
3 --header 'Content-Type: application/json' \
4 --data-raw '{
5 "model": "minimax/hailuo-h3",
6 "channel": "auto",
7 "input": {
8 "prompt": "A cinematic ocean sunrise with realistic waves and smooth camera movement.",
9 "duration": 5,
10 "resolution": "2K",
11 "aspect_ratio": "16:9"
12 }
13 }'1{
2 "code": 200,
3 "message": "success",
4 "data": {
5 "taskId": "task_0123456789abcdef"
6 }
7}codeStatus code. 200 indicates success; other values indicate failure.
messageResponse message. When the request fails, this contains the error description.
data.taskIdApiPass task ID used to query task status and results.
ByteDance
ByteDance
시작가
0 크레딧
ByteDance
ByteDance
시작가
0 크레딧
ByteDance
ByteDance
Turn text, images, video clips, and audio into cinematic, audio-synced videos with a single API call.
시작가
0 크레딧
ByteDance
ByteDance
Generate cinematic 4K AI videos up to 30 seconds with Seedance 2.5's multimodal power.
시작가
0 크레딧
Google Veo 3.1 is the latest state-of-the-art generative video model designed to transform text and image prompts into high-fidelity cinematic visuals. Building upon its predecessors, Veo 3.1 features significant upgrades in prompt adherence, visual realism, and native audio generation, creating 8-second clips with synchronized sound effects and dialogue.
시작가
0 크레딧
wan
wan
시작가
0 크레딧
Runway
Runway
시작가
0 크레딧
Kling
Kling
Transform text prompts and images into high-quality videos with advanced AI. Generate cinematic content with customizable duration, aspect ratio, and audio capabilities.
시작가
0 크레딧
MiniMax
MiniMax
Generate high-quality AI videos from text prompts or images with MiniMax's Hailuo 2.3 model. Support for both 6s and 10s durations, 768p and 1080p resolutions, with intelligent prompt optimization.
시작가
0 크레딧
grok
grok
Grok Imagine is xAI’s AI video generation model for creating high-quality videos with realistic motion, native audio, and synchronized speech. Grok Imagine 1.5 introduces faster generation speeds, improved motion physics, and stronger visual consistency for modern AI video applications.
시작가
0 크레딧
Luma
Luma
시작가
0 크레딧
Kling
Kling
Kling v3 video generation model, supports text-to-video and image-to-video, up to 15 seconds, 1080p pro mode
시작가
0 크레딧
wan
wan
Generate videos with cinematic consistency by reimagining existing footage through high-fidelity visual transformation and structural control.
시작가
0 크레딧
시작가
0 크레딧
grok
grok
시작가
0 크레딧
kling
kling
시작가
0 크레딧