Jimeng AI

What is DreamAI?

Ji Meng AI is an AI-powered platform for creating images and videos, part of the Douyin ecosystem; it is operated by Shenzhen Lianmeng Technology Co., Ltd. Evolving from initial tools for generating images from text or videos, it has become a one-stop \"AI studio\" that offers functions such as idea search, image generation and editing, video creation, intelligent multi-frame handling, a visual canvas, creative agents, the use of digital avatars, community sharing, and the ability to create similar content.

Ordinary creators can use it directly via the Web or app, while companies and developers can invoke models such as Seedance and Seedream through Volcano Engine.

As of this verification, the ImDream app now features Seedance 2.5, which allows for the direct creation of videos up to 30 seconds in length; it also offers functions such as Seedance 2.0 Fast, multi-modal reference, Image 5.0, and Image 4.6.

The names of the consumer-side models for Immediate Dreams are sometimes shown as “Video 2.5” or “Image 5.0”; these may differ from the model identifiers and availability times specified by the Volcano Ark API.

Core functions

• Seedance 2.5 video:The new version allows for the direct creation of videos up to 30 seconds long, making it suitable for short stories, advertising clips, and sequences of actions; the available specifications, resolution, and constraints on materials depend on the page being used for generation.
• Seedance 2.0 Multimodal Reference:It is possible to combine text, images, videos, and audio; new videos can be created by taking into account elements such as the composition of the scenes, details of the characters, their actions, camera movements, rhythm, and atmosphere.
• Multiple material control:According to the official documentation for Seedance 2.0, a maximum of 9 images, 3 videos, and 3 MP3 audio files can be used; the total duration of each video and audio file must not exceed 15 seconds. A maximum of 12 files in total can be uploaded, and the resulting output video can have a duration of up to 15 seconds.
• Video reshooting and editing:It is possible to extend existing videos, connect consecutive shots, and replace, delete, or add characters or elements within the videos. However, complex changes may still affect the consistency of the background and the characters.
• Smart multi-frame:A single continuous shot can be created by using multiple keyframes or video segments; it supports combining videos with each other, merging videos with images, editing specific segments, using only the final frame, and performing hard cuts.
• Image 5.0:Based on Seedream 5.0 Lite, it supports online searching, logical reasoning, and the creation of text-based as well as knowledge-driven images; it can be used to produce posters, social media covers, knowledge graphs, UI elements, and infographics.
• Image generation and editing:It supports text-to-image generation, image generation based on reference images, merging of multiple images, local edits and additions, image expansion, style transfer, and creation of image collections; it can be used for character design, storyboarding, product visualization, and brand materials.
• Canvas and Agent:Multiple image and video assets can be managed within the visualization canvas, and generation tools can be invoked via a ticket system; the Agent offers functions for inspiration search and creative design, understanding tasks through dialogue and organizing the resulting outputs.
• Appearing on camera alongside friends:After recording one’s own image and voice and completing the verification process, users can create a digital avatar and grant friends permission to use it in collaborative creations. It is not a tool for replacing a person’s face without their consent.
• Video 3.5 Pro:Based on Seedance 1.5 Pro, it supports the simultaneous creation of audio and video; it can synchronize the movements of a person’s lips, the sounds of instruments, and background noises, and it also enhances the quality of pronunciation in languages such as Mandarin, Cantonese, and Sichuanese, as well as facial expressions.
• Community and making the same product:Browse other users’ works, view the public prompts, and reuse them with just one click before modifying them to create your own work; the works can be saved locally or shared via link.

Member and point prices

Ji Meng offers a free trial option, membership subscriptions, and the possibility of toping up credits. Different image and video formats, duration, resolution levels, as well as the use of regular or fast processing routes, all result in different amounts of credits being consumed. The rules regarding failed generation, reviews, waiting times, and credit refunds are specified on the task page.

Members receive benefits such as monthly points, extended video duration, and watermark removal; however, this does not mean that all models can be used indefinitely.

According to the official information on automatic renewal, the basic membership plans listed currently are:

• Monthly subscription: 69 yuan per month;
• Continuous annual subscription: 659 yuan per year;
• After activation, member points are earned on a monthly basis, granting access to benefits such as extended video duration and watermark removal.

The App Store in China also lists membership plans for Jimeng at 79 yuan, an advanced Pro subscription at 199 yuan per month, as well as options for top-up with points in amounts of 50 yuan, 75 yuan, 150 yuan, etc. Those products with abnormally high prices marked as “discontinued” on the page should not be considered as current selling prices.

Different membership tiers come with varying numbers of points, VIP fast generation options, resolution levels, as well as different permissions regarding queues and models. It is necessary to check the current benefits in real time before making a purchase; do not estimate the cost based on the information provided in old tutorials regarding free points per day or past deduction amounts.

A recurring subscription will have its payment deducted automatically 24 hours before it expires; to cancel it, it is necessary to proceed with the cancellation through the Apple Account management interface at least 24 hours in advance. Points usually come with specific usage limits and expiration dates, and whether membership privileges, points awarded as part of promotions, or purchased points carry over depends on the guidelines specified for that account.

Volcano Ark API and enterprise pricing

Companies and developers can use Volcano Ark to access the Seedance video and Seedream image models. Dream members do not have any balance associated with the Ark API; the privileges and billing for these two types of accounts are separate from each other.

The video API uses asynchronous tasks, while the image API is billed based on the number of requests and the output specifications; different models may be charged using tokens, the number of images, the duration of videos, or resource packages.

The official package offerings for Seedance 2.0 include: a lightweight creation package with 7 million Tokens for 196 yuan, which allows for the creation of around 28 480P videos; and a comprehensive premium package with 10 million Tokens for 280 yuan, enabling the creation of approximately 40 480P videos.

The package for efficient mass production of 13 million tokens costs 364 yuan, and it includes around 52 videos in 480P quality. The validity period of this resource package is 90 days; it is only suitable for Doubao-Seedance-2.0, not for Fast or Mini versions. It supports videos in 480P, 720P, 1080P resolutions, with durations ranging from 4 to 15 seconds.

The estimated number of generated items is based on the 480P specification; 720P, 1080P, multi-modal input, and longer video durations require more tokens, so it cannot be assumed that 28 tokens will always be sufficient. Once a resource package is used, it is usually not refundable.

Regarding whether Seedance 2.5 and the Jimeng consumer-side VIP model offer APIs, as well as their prices and available regions, the information should be based on the real-time status listed for the Ark model.

Open-source status

The Jimeng AI website, apps, Seedance 2.5/2.0, Seedream 5.0 Lite, as well as the full creative platform are not open-source products; the model weights and training data are not made available by Jimeng for users to deploy locally.

ByteDance has made available on GitHub various research and development projects such as DreamID-V, OneReward, and FlowGram; these may present research results based on Seedance or Seedream, but they do not constitute the release of model weights for consumer use.

Therefore, the catalog should label Immediate Dream as a proprietary cloud service. The official papers, method codes, workflow frameworks, and product models need to be checked separately;

Even if a research repository uses an open-source license, this does not mean that Seedance and Seedream can be used commercially for free or run offline.

Supported platforms

Ji Meng offers both a web version and mobile apps. On the App Store in China, the app is listed as being available only for iPhone, requiring iOS 15.0 or higher; it supports both Simplified Chinese and English.

The web version is better suited for working with canvases, managing multiple assets, and performing detailed edits, while the app focuses on quick creation of content, community features, live recording, and collaborating with friends. Models and events are usually released in batches, and the functions available on the web and in the mobile app may not be exactly the same.

ImDream AI Usage Guide

Complete a basic task.

1. Identify the audience, platform, format, duration, and the information that needs to be conveyed;
2. Prepare scripts, shots, or reference materials that can be used in Jimeng AI;
3. Select the Seedance 2.5 video to create a low-cost preview;
4. Use Seedance 2.0’s multi-modal reference to adjust the visuals, rhythm, subtitles, and audio;
5. Check each frame for characters, text, logos, lip movements, and factual accuracy;
6. Export in the desired format after confirming the licensing for music, portraits, and materials;

Create reusable professional workflows

1. Create scripts, shot lists, brand assets, and a list of elements that are prohibited from use;
2. Save unified parameters for Seedance 2.5 videos, Seedance 2.0 multi-modal references, and multi-material control;
3. First, use representative shots to test the model and the quota;
4. Transfer the failed shots to manual editing or regenerate them;
5. Uniformize subtitles, volume, colors, and end credits;
6. Record the version and reviewer before publishing in batches;

Which users are it suitable for

• Creators who produce short videos, AI-based animated series, advertisements, e-commerce content, and storyboards;
• Designers needed for posters, social media covers, brand visuals, UI, and knowledge graphs;
• Ordinary users who wish to take photos using their own digital avatar or an image authorized by a friend;
• Professional video teams that require multimodal references, video reshoots, and local editing;
• Enterprise developers who use the Volcano Ark to invoke Seedance and Seedream on a bulk basis.

Usage restrictions and security

• Creating AI-generated images and videos requires numerous attempts and trials; errors can still occur with complex actions, hands, identities of people, text, physical relationships, lip movements, and the continuity of long shots.
• The fast track only affects the queuing or generation strategy, and it does not guarantee better results;
• Formal projects should allocate a budget for regeneration, and tasks such as adding subtitles, sound effects, adjusting color grading, and fixing defects should be carried out in video editing software.
• The feature of live video streaming requires identity verification and authorization.
• The current settings of Seedance2.0’s regular material section prevent photos or videos featuring real people from being used as the main reference; users should instead use reliable materials or their own digital avatars.
• It is forbidden to forge others’ statements, pretend to be public figures, or create pornographic or fraudulent content without consent.
• When granting authorization to a friend, it is necessary to specify the scope of the works involved, and to allow that person to manage or revoke the permissions related to the images.
• The app’s privacy policy indicates that it may collect purchase records, phone numbers, photos and videos, browsing history, device identifiers, as well as data related to product interactions and diagnostics;
• Before uploading customer materials, unpublished ads, child-related images, and company documents, it is necessary to verify the authorization and data policies.
• For commercial releases, it is also necessary to verify the copyrights of portraits, music, trademarks, fonts, and reference works, and to comply with the requirements regarding the labeling of AI-generated content.

Frequently Asked Questions

Is DreamAI free?

Some features can be installed and tried out for free, while advanced models, fast track options, and those designed for high usage require a membership or points. The basic membership costs 69 yuan per month on a recurring basis, or 659 yuan for an annual subscription.

How long can DreamAI generate a video in total?

The current version of the app indicates that Seedance 2.5 allows for the creation of videos up to 30 seconds in length; in the multi-modal mode of Seedance 2.0, the typical maximum length is 15 seconds, depending on the model and the interface used.

Does Ji Meng AI support live human appearances?

It supports recording by the user himself/herself, live verification, and digital avatars; it is also possible to take photos together with friends after obtaining authorization. The section for ordinary reference materials imposes security restrictions on real photos and videos.

Does the Dream Member plan include the Volcano Ark API?

It is not included. In other words, the membership points for consumers of the Dream platform and the API of Volcano Ark are subject to two separate billing systems; companies that wish to use these services must acquire them separately and purchase the necessary resource packages, or opt for pay-as-you-go pricing.

Is DreamAI open-source?

It is not open source. The research materials and development repositories made public by ByteDance are not equivalent to the weights of the Seedance and Seedream models, nor to the source code of the Jimeng platform.