Creativity AI #3: Your weekly inspiration of AI-driven innovation and efficiency

Midjourney V7 in Feb, Popular Sref codes, Runway 4K Upscaler, Perplexity O1 model, ElevenReader app, Future transparent AI image and video

Jan 16, 2025

Creativity AI is a new publication that brings together interesting articles and recent developments in the AI world related to creativity and productivity.

Midjourney Office Hours

Due to the delay, Midjourney V7 is now scheduled for release in February.
A bigger batch may be released before Version 7. The final UX or interface is still being worked out.
Due to promising internal tests with updated datasets, the first 3D feature could be released this month or soon.
The new video feature must strike a balance between speed and quality for both casual and professional uses.
Once the new features (V7, video, and 3D) are launched, the free fast /relax mode will come to an end.
Future character references may evolve into Omni-References capable of maintaining consistency (logos and objects).
Most users prefer high-quality (slower) modes and rarely use lower-quality and cheaper modes.

The most popular Style Reference codes in 2024

According to Midjourney, these were the most commonly used Sref codes last year: 698401885, 3322162017, 680572301, 3556861680, 1567159629, 962970979

I used a permutation prompt to test them out.

A minimalist portrait of a woman in profile with a cat perched gracefully on her shoulder, both looking forward. --sref {680572301, 3322162017, 1567159629, 962970979, 698401885, 3556861680}  --v 6.1

Observations and Comments

Midjourney users prefer colorful and brighter aesthetics.
Only one of the six most popular Sref codes has a photographic style. Having said that, I've noticed that most randomly generated Sref codes are biased toward illustration style.
The above Sref codes are now widely distributed in the public domain. If you want to create something more unique and uncommon, consider using Sref codes other than those listed above. (Check out the Boutique Collection and how to protect your prompt).
The code 698401885 was previously shared in the “Free public resources” article.
Midjourney may reveal the popular Sref codes. I assume they had checked to see if those codes were already in the public domain or community before sharing them. If you're creating brands based on specific Sref codes, keep them private to avoid being "overused" and losing their uniqueness.

Niji · Journey

Niji just launched its own moodboard today. It works similarly to the Midjourney Moodboard. You can access it here. The image above was created using the most recent Niji Moodboard.

Niji has released a limited print edition of Niji Journal magazine for $5. But they only ship to the United States, Japan, Korea, and China. If you are lucky enough to live in these countries, order a copy at: https://www.nijijournal.com/store/p/journal-vol-2

Runway

Runway now allows you to upscale videos to 4K resolution. You can find it next to the download options.

After selecting the "Upscale to 4K" option, Runway will create a new 4K video for download.

I tried it and discovered that it also works with previously generated videos. Go to "View all sessions" in the top left corner to find the generation session, and then locate the video you want to upscale.

The upscale to 4K option is hidden in the download menu.

You can find all of your generation sessions in the upper left corner.

Why does this matter?

Previously, users had to use other software to upscale videos in post-production. This new feature saves time and streamlines the production process.

Perplexity

Perplexity has added the O1 model to its list of AI models. Although daily usage is limited to ten, it's quite exciting for users who only occasionally search for information. Daily 10 usages equal 300 searches per month!

Perplexity has partnered with Tripadvisor to simplify travel planning. The Perplexity search results will include hotel images, user reviews, ratings, and other information sourced from Tripadvisor. It has not yet been rolled out to all countries.

Perplexity Pro includes an impressive list of AI models including O1 from OpenAI

Why does this matter?

Perplexity is an excellent platform for searching information. I use it every day. Even more than ChatGPT. The addition of the O1 model is fantastic because it allows me to easily compare the search results returned by various models with a single click.

ElevenLabs

You can create your own podcast on demand using the free ElevenReader App, which is available in the Android and iOS app stores.

Import any content (news articles, eBooks, manuals, journals, PDFs, etc.) into the App and select a voice to read the materials for you while you commute to work, work out at the gym, or walk in the park.

You can even take a photo of something with text and ask the App to read it aloud. I snapped a picture of a label (in French with tiny text) from a candy box, and it read quite well! Impressive!

You can view the introductory video here:

Download the ElevenReader App from the Android or iOS store at https://elevenlabs.io/genfm

Why does this matter?

ElevenLabs offers more AI voices than Google NotebookLM.
It’s free and convenient. You can create the podcast and listen to it on your mobile device.
On mobile, you cannot export the generated podcast or audio clip. Runway charges users credits for text-to-speech on its website. So, if you want to save the podcast as a file for future use, you will need to recreate it.

Tip: If you create your customized AI voice in ElevenLabs, you can select the new voice to read the text through the ElevenLabs Reader App. What's the benefit of doing that? Instead of spending your precious credits to generate new text-to-speech to hear your new character's voice, use the App to have the text read to you for free.

A new method for creating transparent AI images and videos

A Hong Kong-based research group and Adobe have developed a method for improving the visual effects of see-through elements in AI-generated images and videos, such as water, hair, smoke, fire, reflection, and so on. You can read their published article and see numerous examples here.

The examples from the published paper:

Image before transparency — Image before becoming transparent

Why does this matter?

As interesting as it is, the technology has yet to be integrated into Runway or other AI generators. However, given the fierce competition in the AI race, we can expect the technology to become available to the majority of users very soon.

Inspiration

I came across a collection of old fruit images published by Parker Higgins. There are numerous vintage paintings of fruits for the US Department of Agriculture on Parker's Bluesky account. Check out the "old fruit pictures" collection here.

I drew inspiration from it and used Midjourney to create another image of the grapes.