Creativity AI #5: Midjourney V7 Delayed, Perplexity's Android Push, and DeepSeek's Rise
DeepSeek, Janus-Pro-7B, Midjourney Office Hours, Free Midjourney resources, Describe Image, Perplexity, and ElevenLabs
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 (2025-01-29)
The Midjourney V7 release is expected to be delayed by another week due to unexpected issues.
The V7 will feature improved character and object references (logos, mugs, animals, etc.). The goal is to have all references and functionality ready for the launch.
The large batch (Batch 8) is in development, but the priority right now is to get V7 ready for release.
The new video feature will be a key Midjourney offering this year.
The 3D feature will also be released this year, but the timeline is unknown.
Midjourney is working on hardware research and other "secret projects" to maximize the use of its hardware and software expertise.
Free Midjourney Resources
Midlibrary has just released 2,500 Sref codes available for free in the public domain. You can browse and select the codes you want here.
Previously, the website was listed as a free public resource for the Midjourney community. Read it here to find out about other free resources.
Midjourney Updated “Describe Image” (/describe) on the Website
In addition to the existing Describe feature on Discord, there is another one available on the web. For details and secret tips on how to use the new "Describe Image" feature, read this article (paid members only).
Perplexity
Perplexity launched Perplexity Assistant for all Android users. It is an excellent addition that significantly broadens Perplexity's use cases beyond the desktop.
Read the full article here.
Perplexity has also added DeepSeek R1 to its list of AI models.
Why does this matter?
The Assistant has made using Perplexity on mobile second nature. You can ask the bot any question simply by speaking. The Assistant can access the camera, analyze what is on the screen, launch apps such as Google Maps and Spotify, and even send emails and reminders.
Take full advantage of your Perplexity Assistant search function. Bring AI into your daily life.
Perplexity users can now use the DeepSeek model without having to install it locally on their computer or mobile device.
ChatGPT
ChatGPT launched Operator - an AI agent for the Pro users ($200/ month). It can browse the internet, clicking around the web page to make a reservation in the restaurant, buying tickets and so on, just like a human using a web browser - according to OpenAI.
While the bot can simulate the use of browser to accomplish tasks, I’m unsure if the Operator alone worth $200/month subscription. I do not subscribe to Pro so unable to compare it with Perplexity Assistant ($20/month) that appears to accomplish more or less what Operator could do except simulating the use of browsers.
Why does this matter?
The competition for the AI Assistant market has already begun. Just one day after Perplexity launched its Assistant, Google Gemini aggressively sent out messages to Android mobile users, persuading them to make Gemini their default Assistant. It's interesting to see who will win and take over the market.
Are you using AI Assistant or OpenAI's Operator? Please let me know what you think about this new feature in the comments.
ElevenLabs
ElevenLabs has released a minor update: "Projects" has been renamed Studio, and it now includes a slash menu where you can type / anywhere in the text to add a pause (up to 3 seconds).
The text is now highlighted when the bot reads it, making it easier to follow along.
Why does this matter?
A quality-of-life improvement to add pauses in AI-generated voices. Improve the overall experience of content creation.
DeepSeek
It was a DeepSeek week, the latest shining thing in the AI world, which was widely covered in the news and on social media and became the most downloaded free app on Apple's App Store.
DeepSeek's success has been attributed to its advanced AI capabilities achieved with fewer development resources, open source, as well as its API's lower cost when compared to other more established AI models such as ChatGPT and Claude.
But what's in it for me as an end user when compared to other options?
I tried DeepSeek.
Here is what I found:
Pros:
Free
It’s now available on Perplexity
It shows its thinking and reasoning process
It works somewhat like other AI text models, with a simple interface
The disadvantages:
No voice chat function
No AI assistant
No custom GPT
No image creation (but it announced its image generation model, named Janus-Pro-7B - see below)
No automatic reference of the sources
Why does this matter?
The DeepSeek low pricing model serves as a wake-up call to other AI companies. It is possible to develop AI solutions at a lower cost. And please don't charge us $200 for a monthly subscription.
Rather than jumping right to the answer, DeepSeek reveals its thought process to users, which is a refreshing and interesting feature.
DeepSeek may have limited features right now, but this is most likely due to its early stage. Who knows what features it will develop to truly compete with or even outperform other AI companies.
The following is an excerpt from DeepSeek's thinking process. It's fun to read for the first few enquiries, but it can be tiring if you need an answer quickly.
Okay, so the user is asking who funded the development of DeepSeek. Let me start by recalling what I know about DeepSeek. From what I remember, DeepSeek is a Chinese company that focuses on AI research and applications. They've developed various AI models and products, similar to other AI companies in the space.
First, I need to confirm whether the funding information is publicly available. I know that many tech startups, especially in AI, often receive venture capital funding. Maybe DeepSeek was funded by venture capital firms. Let me think if there were any notable announcements or press releases about their funding rounds.
I should check if there were any specific VC firms mentioned in relation to DeepSeek. Sometimes companies announce Series A, B, etc., funding. Alternatively, they might have been incubated by an accelerator or received government grants, especially since they're based in China.
Wait, I recall that DeepSeek might have been backed by major Chinese tech investors. Maybe companies like Tencent, Alibaba, or Baidu? Or perhaps Sequoia Capital China, which is active in tech investments. Alternatively, maybe it's a spin-off from a university or research institution, which sometimes get funding through academic or government channels.
… (truncated)
Given that I can't recall specific funding details, I should state that the exact funding sources aren't publicly disclosed in widely accessible sources. However, it's common for such companies to be funded through venture capital, private investments, or possibly government grants, especially in the AI sector. Advising the user to check recent press releases or official statements from DeepSeek for the most accurate information would be prudent. Also, considering the dynamic nature of startups, funding details can change, so the latest info would be best obtained directly from the company or reliable news sources.
Tip: Perplexity will automatically hide the DeepSeek’s thinking process in the collapsible panels. However, you can click to expand it for reading.
DeepSeek Janus-Pro-7B
You can test the DeepSeek text-to-image generation model here. It's located at the bottom of the page. The image resolution is 384 x 384.
You can only generate a few free images before the system requests that you pay for GPU resources.
Why does this matter?
Janus-Pro-7B is still in its early stages and cannot compete with more established text-to-image models such as Midjourney and Ideogram. But who knows what it will evolve into in the future?
David Holz of Midjourney stated that it is encouraging to see new innovation (in the image generation space), and that AI companies ought to keep up their development pace.
Janus-Pro-7B is not yet suitable for the AI art generation—more of an experiment for people to try out.
Inspiration
Inspired by photographs of fireflies in the forest, I used the phrase "hypnotic fireflies" to create the prompts and images below. The keyword adds an element of magic to the scene.
Recent articles
Free Geeky Animals' Sref code: 3896264700
Cover prompt: A fashion photograph of a strong man with a sharp jawline, standing tall amidst stormy winds, holding an ancient scroll that glows faintly, signifying wisdom and perseverance. --ar 16:9 --sref 3896264700
Invitation to publish on Creativity AI
Please contact me if you want to publish on Creativity AI.
This publication will be distributed through Substack and Medium platforms, reaching thousands of readers.
Email: animalsgeeky@gmail.com