3-Month Plan | Midjourney Office Hours Recap Jan 10th 2024 | Midjourney News

Summary

The transcript summarizes a recent Midjourney office hours video update. The main topics covered include updates on Midjourney version 6, new website features, goals for the next few months, and potential upcoming capabilities. The team is working on follow-ups and new features for the version 6 launch. They pushed an update to improve aesthetics, coherence, text rendering and stylization. However, they need more servers to make version 6 the default, as current capacity cannot support all users. Inpainting training will finish this week but release timing is uncertain - maybe end of this week if lucky or end of next week if not. Inpainting will allow zooming out and panning. They also updated content filters. Version 6 will be tested with guides and mods next week. It may not initially support text input due to language handling complexity. A timeline feature is slated for the website to help synchronize code bases and ease updating. They may expand the creation alpha soon, pending usage and enjoyment. The core create page needs reworking and explore page development is slower than desired. They are considering AI filtering of bikini images to prevent undesirable content. The January goal is getting necessary features on the website. February will involve polish for better flow. March goals are more user access and new onboarding without a Discord requirement, GPU permitting. Major upcoming capabilities involve research on consistent character and style. A style feature may ship this month and basic character capabilities later, though timelines are uncertain. Video is targeted for February/March - aiming for high quality upscaling that maintains artistic integrity. They will train models on pose and composition control nets this month but results are unpredictable. The new text describe feature is deprioritized for now. Example art was created using a prompt on Mobius style sci-fi fantasy art with details and cinematic composition.

Takeaways

• They are working on follow-ups and improvements to the Version 6 website launch
• They pushed an update to the default aesthetic along with better stylize values, coherence, and text rendering
• Inpainting training is still ongoing, estimated release is end of this week or next
• Version 6 will be tested with guides and mods next week
• They are working on synchronizing code bases between websites for easier updating
• They may add an AI filter to stop inappropriate bikini images from appearing publicly
• Their goal for January is getting all necessary features on the site, then polish in February and expand access in March
• Major upcoming features involve research into consistent character and style
• Video feature planned for Feb/March will prioritize quality over quantity
• New 'control net' feature for setting pose and composition is in early stages

Q & A

What improvements are planned for image rendering?

Better stylize values, coherence, and text rendering

When will inpainting be publicly released?

Estimated for end of this week or next week

How will they test new features before public release?

By having guides and mods try Version 6 next week

What will a potential AI filter target?

Blocking inappropriate bikini images from appearing publicly

What is the development roadmap for the next few months?

Get necessary features in January, polish in February, expand access in March

What major features are they researching?

Consistent character and style

How will the upcoming video feature compare to existing generators?

It will prioritize quality over quantity of results

When could the new 'control net' feature launch?

No guarantee on release date yet, still in early stages

What code base work are they doing?

Synchronizing between sites for easier updating

Who will test Version 6 first?

Guides and mods will test it next week

Outlines

00:00

Midjourney Team Updates

The Midjourney team provided updates on recent progress and future plans. They pushed updates to the default aesthetic and text rendering in Version 6. In-painting training is underway and may release this week or next. They plan to test Version 6 with guides and mods next week. They are working on timeline, creation page, explore page, and onboarding improvements. Major upcoming features involve research into consistent character and style.

Potential Video and Control Features

Video upscaling that makes images move slightly may release in February/March. They are training models this month for potential control features like posing and composition, but results are uncertain. The new 'describe' feature is on the back burner for now.

Prompt and Art Sharing

The narrator shares a prompt for creating art in the video style. They invite viewers to follow them on Instagram and support them on Patreon for more prompts and art. They thank Mike Linden for the prompt found on Midjourney's Explore page.

Keywords

💡Version 6

Version 6 refers to the next major update to the Midjourney AI art generation platform. It will include improvements to image quality, new features like zoom and pan, upgraded safety filters, and expanded server capacity to handle more users. The launch has been delayed due to capacity constraints but is slated for late January or February 2024.

💡Inpainting

Inpainting is an AI technique for filling in missing or corrupted parts of an image to make it whole. Midjourney is training new inpainting models that will allow users to erase and edit objects in their generated images. This long-awaited feature will provide more artistic control but the release timeline is still uncertain.

💡Explore Page

The Explore page surfaces interesting and trending AI art from the Midjourney community. David mentioned wanting to overhaul this page, likely to improve discovery of top creations and prompts.

💡Onboarding

Onboarding refers to the process of introducing and teaching new users how to use the Midjourney platform. They aim to improve onboarding in March 2024 to smoothly ramp up new users without Discord accounts.

💡AI Safety

As an AI art leader, Midjourney is considering societal impacts and potential risks. The script mentions filters to block inappropriate bikini images as an example of their approach to safety.

💡GPU Capacity

GPUs are the specialized computer chips that power AI image generation. Expanding Midjourney requires scaling GPU resources, which is challenging with supply shortages. This capacity constraint has slowed feature development and new user onboarding.

💡Style Consistency

Style refers to the aesthetic look and artistic qualities of an image. Consistent style allows users to maintain the same style across a set of images. This advanced capability reproduces an art style more reliably and is under development.

💡Video Generation

The next frontier for Midjourney is video creation using AI. Initial capabilities to animate images and produce short videos may arrive by February 2024, with quality and reliability improvements continuing over time.

💡Control Nets

Control nets allow steering the AI art generation process via user inputs like sketches, poses, and compositions. Though still experimental, control nets aim to give creators more direct and precise artistic direction over the outputs.

💡Describe Feature

The describe feature in Midjourney takes a text description and generates corresponding images. Progress has apparently stalled recently as the team is not satisfied with the current iteration. This feature remains in flux.

Highlights

They pushed an update to the default aesthetic along with better stylize values, text rendering, and coherence.

Inpainting training is happening this week, but the release date is unknown. It may include zoom and pan.

Version 6 will be tested with guides and mods next week. Text drawing support is complicated.

They want to add timelines and synchronize code bases to allow easier website updates.

They may expand the creation alpha soon if people are enjoying it.

The core create page will likely be redone. The explore page development is slower than desired.

Their January goal is getting necessary features on the site. February is polish and flow. March is onboarding without Discord.

Major features in development are consistent character and style. A style feature may ship this month.

Basic character features may come in Feb/March. Video is aiming for upscaling with high quality.

They will train some models for control net this month but results are uncertain.

The new describe feature is on the back burner for now.

Follow FutureTechPilot on Instagram for MidJourney pictures.

Support FutureTechPilot on Patreon for exclusive prompts.

This recap covered the MidJourney office hours and development updates.

MidJourney is actively improving the platform with new features and fixes monthly.

Transcripts

hey everyone here's a recap of the mid

Journey office hours from January 10th

they are working on follow-ups to the V6

launch and getting as many features as

possible on the new website last weekend

they pushed an update to the default

aesthetic along with better High stylize

values better coherence and better text

rendering they're still looking at more

improvements specifically looking for

Server updates right now if version 6

was the default they wouldn't have

enough servers for everyone they're

working on in painting in fact it's

training right now and it will be done

training this week but they don't know

when the official release will be ready

if they're lucky end of this week if

they're unlucky end of next week in

painting will also include zoom out and

Pan the weird parameter was also fixed

recently and they updated they're not

safe for workplace filters so we should

get fewer unintended nudity nii version

6 will be tested with the guides and

mods next week it is possible it won't

support drawing text apparently there's

something complicated about handling

multiple languages but the team wants to

accomplish it so we'll have to wait and

see David said something about a

timeline being added to the website and

I think that's referring to the alpha

only he said they're working on

synchronizing the code bases a little

more so that each website is able to be

updated more easily and they will

probably expand the creation Alpha soon

right now they're looking at are people

even using it and if people aren't using

it or enjoying it does it make sense to

allow more people to use it that's kind

of the point they're at right now he

said he isn't happy with the core create

page I'm not really sure what he meant

by that but he did say it will probably

get redone he also isn't happy with the

explore page development is moving

slower than they want he did mention

that he's thinking maybe you could make

bikini images but there could be an AI

filter that will just stop those images

from showing up on the front page that

might be a good workaround their goal

for January is to get all the features

that are necessary onto the website then

in February he said they want to polish

it so it all flows better and in March

giving more people access as well as a

new onboarding process basically being

able to use mid Journey without a

Discord account however that is slightly

dependent on GPU availability they

really don't have much more room for

more users in fact open AI had the same

issue recently and had to stop signups

then he mentioned some stuff with we can

look forward to he was talking about

major major features involving current

research and in that sense he was

talking about consistent character and

style they have a good thing for style

that they might ship this month they're

working on character stuff and maybe

they might have a basic feature they can

release along with style he said that he

doesn't want an early version that will

then be changed in the future he would

rather have a solid complete version of

a feature but he knows that's not

entirely realistic then David said maybe

February or March for video and it will

be like an upscale that makes the image

move a bit they're aiming for high

quality and a high probability that you

like the generation rather than the more

Miss than hit you get with other video

generators at the moment he talked a

little bit about control net as in

setting a pose composition maybe a

drawing input and they're going to train

some models this month but no guarantee

on what the results will be like

unfortunately the new describe feature

might on the back burner I just don't

think he's happy with what they

currently have if you like the art in

this video here is the prompt Mobius art

style 80 sci-fi fantasy magazine art

professional illustration hyper details

stunning image of a subject action

background cinematic composition Das

Dash no fake CGI I used Chaos 2 stylize

60 100 and style raw for some of these

big thanks to Mike Liden over on the mid

Journey website I found this prompt on

the explore page if you want a better

look at some of my pictures you can

follow me over on Instagram at Future

Tech pilot if you want some exclusive

prompts you can support me over on

patreon there's your recap of the mid

Journey office hours I hope you're doing

well take care and I'll see you next

time