Midjourney

What is Midjourney?

Midjourney is an AI image generation platform renowned for its ability to deliver aesthetic results and precise style control. Users can create illustrations, photographs, conceptual designs, and commercial visuals by using textual prompts, image references, and various parameters. It is currently used primarily on the Create and Edit pages of the Midjourney official website, with a Discord Bot workflow also available.

The product allows for the creation of 5-second short videos from images, with a possible extension to up to 21 seconds. All plans are based on automatic subscription renewal; there are no long-term free options. The key resource that needs to be purchased is the GPU time required for the model to process tasks.

Main functions of Midjourney

• Generate images in various styles and aspect ratios using textual prompts;
• Upload images as a reference for content, composition, or visual direction;
• Use Style Reference to reuse colors, materials, and overall style;
• Use Omni Reference to improve the consistency of characters or objects;
• Parameters are used to control the version, level of stylization, randomness, and aspect ratio of the image;
• Use Variations to create similar compositions or greater variations;
• Change the canvas range by using Zoom Out, Pan, and zooming in;
• Restructure, erase, partially redraw, and adjust layers in the web editor;
• Use Describe to get inspiration for prompts from uploaded images;
• Convert generated images or existing pictures into short videos with low or high motion.
• Set the starting frame, ending frame, repetition setting, and the number of videos in a batch;
• Manage works through Organize, folders, search, and ratings.

Web page creation and Discord

The Create page on the official website offers visual prompts, a parameter panel, options for selecting reference images, and a history of created works – making it suitable for new users. Discord uses commands such as Imagine to enable quick creation of content, easy management of parameters, and viewing of queues within channels; however, the content in public channels is more accessible to others.

Image reference and style control

A regular Image Prompt affects the content, composition, and colors, while a Style Reference is used to transfer the visual style. Omni Reference is employed to specify certain people or objects. The reference images serve only as a guide, not as a basis for pixel-by-pixel replication; excessive use of them can limit creativity and may pose risks related to trademarks and copyrights.

Web image editor

On the Edit page, it is possible to upload external images, adjust the canvas, move or scale elements, and redraw specific areas using masks. This editor is suitable for fixing certain parts of an image, changing the background, and enlarging images; however, detailed text, precise logos, as well as the structure of products still require review using professional image editing software.

AI short videos

All packages allow video generation in Fast mode; the initial length is 5 seconds, with an increase of 4 seconds each time, up to a maximum of 4 increases, resulting in a length of 21 seconds. Basic can only produce SD-quality videos, while Standard, Pro, and Mega also enable HD-quality video generation in Fast mode.

Only Pro and Mega support generating SD videos in Relax mode.

By default, the video generation process produces 4 results at a time; the value of Batch Size can be set to 1, 2, or 4 in order to control the GPU usage costs. Currently, generating 1, 2, or 4 images in SD format takes approximately 2, 4, and 8 minutes respectively, while generating such images in HD format takes around 7, 13, and 26 minutes respectively. The cost of generating additional images is the same as the cost of the initial generation.

Fast, Relax, and Turbo modes

The monthly package for Fast includes a certain amount of GPU time, and any unused time within this quota is not carried over. Relax places tasks in a dynamic queue without using up the Fast quota, making it suitable for image processing tasks where there is no deadline.

Turbo is usually faster, but it consumes about twice as much GPU power as Fast; moreover, some of the latest model versions may not be supported by it.

Midjourney prices

Midjourney offers four tiers: Basic, Standard, Pro, and Mega. Payment is made annually, which gives a 20% discount compared to monthly payment. With all plans, it is possible to purchase Fast GPU time at an additional cost of $4 per hour.

Concurrency, video, and batch processing capabilities

How to choose a package

Midjourney usage guide

Generate the first set of images from text.

1. Log in to the official website and verify that you have purchased a subscription;
2. Go to the Create page and enter the subject, environment, and action in the prompt box;
3. Meet the requirements regarding supplementary media, lighting, lenses, and colors;
4. Set the appropriate aspect ratio and model version;
5. First, generate a set of results with a moderate level of stylization;
6. Select the best image for Variation or local editing;
7. Zoom in to examine text, hands, edges, and brand elements;
8. Save prompts, parameters, and generation records before downloading.

Use the reference diagram to maintain the style.

1. Confirm that you have the right to upload reference images;
2. Select Image, Style, or Omni Reference depending on the purpose;
3. Avoid putting reference diagrams with different purposes in the same slot.
4. Use text to clearly indicate the parts that must be retained and those that can be changed;
5. Start testing with a lower weight;
6. Compare multiple results and gradually increase or decrease the weights;
7. Finally, verify the identity of the individuals, the logo, the product structure, and any copyright-related risks.

Convert images into short videos

1. Choose a image generated by Midjourney or upload your own picture;
2. Verify that the screen aspect ratio matches that of the final publishing platform;
3. Choose Low Motion or High Motion;
4. To save GPU resources, set the Batch Size to 1 first;
5. Enable Loop when a loop is needed, or add an end frame;
6. Generate SD samples first, and only consider HD after confirming the direction of movement.
7. Increase the duration by 4 seconds each time and check the continuity of the transitions;
8. After downloading, use video software to edit the audio, subtitles, and carry out the final editing.

Who is Midjourney suitable for?

• Illustrators and concept designers: exploring composition, color, and visual direction;
• Brand and advertising team: Create mood boards, campaign concepts, and visual drafts;
• Game and film production teams: designing characters, scenes, props, and reference frames for storyboarding;
• E-commerce and social media management: rapid creation of backgrounds, posters, and short video content;
• Architects and interior designers: explore different styles and discuss design concepts;
• Independent creators: creating covers, avatars, story illustrations, and experimental videos;
• Researchers: Exploring the aesthetics and interaction methods of generative visual models.

The advantages of Midjourney

• The overall aesthetic, lighting, and stylistic elements of the generated image are consistent;
• The combination of text, images, style, and subject references is quite flexible;
• The web interface has reduced its reliance on Discord commands;
• Variation, image enlargement, local editing, and video creation form a continuous workflow;
• Above Standard tier, unlimited Relax images are provided;
• Pro and above offer Stealth and Relax SD videos;
• The official documentation provides clear explanations regarding GPU processing time, concurrency, and the differences between plans.

Usage restrictions and precautions

• There is no long-term free version – all plans are automatically renewed.
• If the Fast GPU capacity is not exhausted, it will not be carried over, and processing videos takes more time than processing images;
• “Infinity Relax” is still subject to constraints related to queues, concurrency, and fair usage.
• Errors can still occur when generating text, precise logos, hands, and complex structures.
• Stealth is available only in Pro and Mega versions, and it cannot be used to make public Discord channels private.
• To upload real people, trademarks, and copyrighted works, one must possess the corresponding rights;
• The generated content may exhibit similarities to existing works, characters, or brands;
• Before commercial release, manual design as well as copyright and platform policy reviews must be carried out.

Privacy, rules for making works public, and commercial use

By default, Midjourney is a community-oriented platform, and the works created there can appear in the exploration section of the website as well as in public galleries. The Pro and Mega versions allow users to hide their works from the official website, but content created in public Discord channels will still be visible to those members of the channel; for projects that require confidentiality, it is necessary to use web pages or private messages and ensure that privacy settings are in place.

Subscribers can typically use the generated images and videos in accordance with the standard commercial terms, but companies with an annual revenue of over 1 million dollars must purchase Pro or Mega. Users should ensure that they have the rights to use the content they upload, and be aware that the copyright protection for AI-generated content may be limited in different jurisdictions.

API, GitHub, and open-source information

Midjourney does not provide an official generation API available to ordinary developers. Numerous so-called Midjourney APIs available online work by automating Discord functions or collecting user tokens; these are unofficial solutions that may violate the service terms and pose risks to accounts and keys.

The official GitHub organization makes available the source code of older versions of the documentation, as well as certain research-related dependencies and branches; however, the core image and video models, training data, and production services are not made available under an open-source license. The presence of these public repositories does not mean that it is possible to download Midjourney’s weights or to deploy the entire platform on one’s own.

Frequently Asked Questions

Is Midjourney free?

There are no long-term free plans available at the moment; the lowest Basic plan costs $10 per month or $96 per year. Automatic renewal is enabled by default for all plans.

What is the difference between Standard and Pro?

Standard includes 15 hours of Fast mode and unlimited Relax mode images; Pro offers 30 hours of Fast mode, as well as Stealth mode, more simultaneous connections, and SD videos in Relax mode.

Can Midjourney generate videos?

It is possible to create a 5-second video from an image, with the duration increasing by 4 seconds each time, up to a maximum of 21 seconds. Plans above Standard allow for HD-quality videos, while Pro and Mega enable SD-quality videos in Relax mode.

Can Midjourney creations be used for commercial purposes?

Subscribers can usually use the service under standard commercial terms. Companies with an annual revenue of over 1 million dollars need Pro or Mega, and they still have to deal with risks related to trademarks, portraits, and copyrights.

Is Midjourney open source?

No. The documentation and research repositories made available on the official GitHub do not include the complete generation models or production platforms.