Aud-API

Overview

Introduction

The Aud Open Platform currently provides the following capabilities to empower developers:

Text-to-Video: Input a text to generate a video; currently supports general style as well as specifically optimized anime style;

Image-to-Video: Input the first frame image or both the first and last frame images along with the corresponding text description to generate a video;

Reference Video Generation: Input reference images and text descriptions to generate videos. Currently, it supports multiple types of subjects such as reference characters and objects;

Intelligent Ultra HD: Input a video, and generate a video after super resolution and frame interpolation (fixed to 1080P, 24 frames). Currently, it only supports super resolution for videos generated by the Aud model.

Application scenarios

Application scenarios supported by the Aud Video Generative Model include but are not limited to:

Social media interactive gameplay: including two-person hug, old photos come to life, personalized emojis, etc.;

Production of short video content such as animations and games: Generate video clips based on the storyboard and edit them into a complete video;

Virtual IP Daily Interactive Promotion Content Creation: Production of IP promotion short videos and generation of interactive videos;

E-commerce Ad Production: Generate empty shot creatives for ads, create special effect shots based on static images, etc.

Advantage Description

Large dynamic range and imaginative: The generated video has large action amplitudes, avoiding the slowness of AI PPT, and is highly suitable for presenting imaginative visuals or special effects gameplay;

Powerful semantic understanding: It can accurately follow the input prompt instructions, especially the image-video capabilities can generate various images according to the prompts, making card drawing more efficient;

Outstanding anime-style effects: Text-to-anime-style video generation produces better results, and when using anime images for image-to-video generation, the style stability is high, with few issues such as style changes;

Stable subject reference effect: It overcomes the consistency issue that is difficult to solve at the image level, can achieve high consistency of the subject, and can generate videos with consistent subjects without the need to create a first-frame image, making creation simpler.

Product Pricing

The Aud Open Platform provides rich video generation API capabilities to meet the diverse needs of enterprises and developers. Currently, API calls are priced using a point-based system.

Points	Price (RMB)
1	0.3125

Optional coin package prices

For business customization and bulk purchases, please contact xxxx.xxxxx@ .cn . We will serve you wholeheartedly

Amount (RMB)	Number of points purchased	Available concurrency count	Valid Period
500	1600 points	≤ 5	1 year
1000	3200 points	≤ 5	1 year
2000	6400 points	≤ 5	1 year
5000	16000 points	≤ 5	1 year
20000	64000 points	≤ 5	1 year

Precautions:

Videos that fail to generate (including failures caused by content moderation rejection) do not incur credit consumption;

Available concurrency ≤ 5, which means that at most 5 concurrent tasks can be used. In rare cases, system load may cause your concurrency to be lower than the maximum value.

Points Consumption Details

Video generation capabilities

Model Version	Capability	Duration	Clarity	Style	Points Consumption	Off-peak point consumption
Aud Q1	Reference - Video	5S	1080p	Universal	8	4

Audio generation capabilities

Capability	Model Version	Duration	Points Consumption
Text-to-Audio	audio1.0	Less than 5 seconds	1
Text-to-Audio	audio1.0	Less than 10 seconds	2
Controllable text-to-sound effect	audio1.0	Less than 5 seconds	1
Controllable text-to-sound effect	audio1.0	Less than 10 seconds	2

Scene Template

We provide a variety of high - popularity and stable - effect scene templates such as “Hug”, “Kiss”, and “All Things Blossom”, and they are continuously being updated; different scene templates have different rules for points consumption;

Fine-tune

The price of fine-tuning the model is related to the Batch Size and Step you choose, and the consumption details are as follows

Batch Size	Step	Required Points	Estimated Time/Hours
4	1000	20000	8
4	2000	40000	14
4	3000	60000	18
4	4000	80000	24
8	1000	40000	14
8	2000	80000	24
8	3000	120000	36
8	4000	160000	50

The trained fine-tuned model, consumes 4 credits per inference call .

The current fine-tuning feature is in the internal testing phase. If you would like to experience it, please send an email to xxxx.xxxxx .cn to apply for permission.

Recommended prompts

Billing is based on the number of output prompts. For every 5 output prompts, 1 credit is deducted.

lip sync

The price of lip-syncing is related to the duration of the video you generate, and the consumption details are as follows

Capability	Points consumption per second
lip sync	1

Frequently Asked Questions

Aud API Usage Issues

#	Question	Answer
1	How to purchase and activate API services?	After completing registration and login, you can recharge and purchase the credits required for generating videos in the “Account Recharge” menu on the platform.
2	How to make an API call?	You need to use the “HTTPS” protocol and the “POST” method to call our API; place the API key in the “Authorization” field of the request header, and include the parameters required for video generation, such as model, style, duration, resolution, etc., in the request body. The specific parameters and the domain name for the call can be found in the API documentation.
3	How to obtain an API key?	After completing registration and login, you can create a new API key under the “API Key Management” menu on the platform.
4	What if the API call fails?	If the API call fails, you can first check whether the network is normal and whether the parameter settings of the request headers and request body are correct. If the problem persists, you can fix it based on the error information and prompts returned in the response body. Or contact xxxx @xxxx.cn We will serve you wholeheartedly.
5	What is the maximum number of characters supported for prompt input?	Supports up to approximately 1500 characters.
6	Which languages does the prompt support?	Currently, we support over 200 different language inputs worldwide.
7	Can the optimized result of the prompt be returned?	Currently, returning the optimized results of prompts is not supported.
8	What are the ways to upload images?	Supports passing in Base64-encoded data of the image or the image URL. If using a URL, please ensure the resource is publicly accessible; if using Base64 encoding, please ensure the data format is correct and complete.
9	What formats are supported for the input images?	Supports JPG, JPEG, PNG, and WebP formats.

Issues related to the Aud model

#	Question	Answer
1	Aud What video generation durations does it support?	Supports two video generation durations of 4 seconds and 8 seconds, where the 8-second duration is only supported at 720p resolution.
2	Aud supports which resolutions?	Supports three resolutions: 360p, 720p, and 1080p; and for videos with a generated resolution of 360p, intelligent ultra-high definition processing can achieve 1080p.
3	How long does it take to generate a video?	Generally speaking, generating a 4-second video with a resolution of 360P takes about 10 seconds, and the time doubles for an 8-second video. The reasons affecting the generation time are relatively complex, specifically including: input prompt and image complexity, review session duration, concurrency limit, and interface response efficiency, etc.

Authentication mechanism

How to Obtain Authentication

To ensure the legitimacy of the requester’s identity and that the request is not maliciously tampered with during transmission, < Aud >’s signature mechanism requires the requester to construct a signature and then send the signature along with the API request to the server. The server will calculate the signature of the received request using the same mechanism and compare it with the signature sent by the requester. If the signature fails verification, the request will be rejected. Developers need to complete the following operations according to the process:

GetSecret Access Key

Create a standard request

Create string to be signedstring

Construct Signature

Add signature to request

The authentication steps for different products may vary. You can modify the content in the above process and the corresponding titles in the following text according to the actual situation.

GetSecret Access Key