EP 045 · 75 MIN A CHILDHOOD DRAWING CAME TO LIFE

Midjourney Video Breakdown Part 2

July 6, 202575 min Rory Flynn & Drew Brucker

Watch on YouTube Spotify Apple Podcasts Audio file

The short version

Episode 45 is a two-week check-in on Midjourney video, where Rory finds 480p often looks more coherent than 720p or 1080p and shows a 'transformative prompting' trick that morphs an abstract skull image into a ballerina. The hosts test wild style-code prompts like 'unglued photography,' animate a childhood drawing Rory's mom saved, and build a quick Weavy tool to extract a single frame from a generated video. They also cover roadmap news on sound and a pricier pro video tier.

Key takeaways

01
480p often looks more coherent in Midjourney video than 720p or 1080p, since less pressure on sharp detail lets the model focus on motion; upscale after with Topaz.
02
Simple keyword prompting (old V5-style) still works for Midjourney video: mood words like 'peace and serenity' or 'aggressive' reliably shift the whole aesthetic.
03
Let Midjourney auto-generate a video first (high and low motion) before manually prompting: its first pass often suggests directions you wouldn't have written yourself.
04
Pick your weirdest, most abstract Midjourney images for Midjourney video specifically: Rory says those are the ones no other video generator could handle anyway.
05
Rory built a 4-node Weavy flow in about two minutes to extract, crop and upscale a single frame from any generated video.
06
Midjourney is considering ambient sound effects (wind, footsteps, traffic) before speech synthesis, since mismatched dialogue is harder to forgive than missing background noise.
07
Midjourney's business API will likely ship trimmed down, Rory predicts, pointing to Kling's API as precedent: full-featured on the website, just 'generate video' via API.

Receipts · said on the record

“If it's mismatched footsteps of someone running like you, your brain immediately picks up on that and says that sucks.”

Rory Flynn Why one mismatched sound effect breaks an otherwise good AI video

“I I will gladly sacrifice and utilize 480p if I get … the motion or the effect that I'm looking for, and then I can upscale it with Topaz, which has been great.”

Rory Flynn Why Rory prefers 480p over higher resolutions for Midjourney video

“This was a picture my mom sent me of an image that I drew her when I was like seven or something … she has a wooden spoon that's shooting a rainbow at a at a dragon.”

Rory Flynn Rory animates a childhood drawing his mom saved, using Midjourney video

“Imagine just the context switch in the medium: if you're a kid, you draw something and then all of a sudden you do something like that, and it's just like, yo, these are novel things happening here.”

Drew Brucker Drew on why AI-animating a childhood drawing feels so magical

“It's almost like the PLG motion behind it. It's just like, hey, just get the product in people's hands first. And then you, you know, people will justify the spend after too.”

Drew Brucker Drew on Midjourney's pricing strategy: get the product in hands first

Chapters

Questions this episode answers

Does lower resolution actually look better in Midjourney video?

Sometimes, according to Rory: he found 480p often reads as more coherent than 720p or 1080p, because the model isn't as focused on sharp detail and can prioritize motion and coherence instead. He upscales the result afterward in Topaz.

What is 'transformative prompting' in Midjourney video?

It's Rory's term for prompting a video model to morph one subject into something else entirely, such as turning an abstract skull-and-butterfly-wings image into a ballerina, with a seamless transition rather than a hard cut.

Is Midjourney adding sound to its video feature?

The Office Hours notes discussed on the episode suggest Midjourney is considering ambient sound effects, like wind, footsteps or traffic, before attempting full speech synthesis, since Rory argues mismatched dialogue is far more noticeable and damaging than missing background sound.

What will Midjourney's business API look like?

Rory predicts it will ship with a trimmed feature set, pointing to Kling's API as precedent: full of features on its own website, but reduced to something like plain 'generate video' once accessed via API, likely without style references or permutations at first.

How did Rory extract a single frame from a Midjourney video?

He built a four-node flow in Weavy, in about two minutes, that uploads a video file, extracts a chosen frame, adjusts its composition, and upscales it, which he then ran through Magnific.

Show notes

Fast Hours Podcast Ep. 45—Midjourney Video Breakdown Part 2

Rory and Drew are back and things get... weird.

In this episode, they push Midjourney Video to its limits—trying to animate mushrooms, morph soda cans, and see what happens when you feed it your childhood art.

Along the way, they break down the newest Office Hours updates, upcoming pro features, and the mystery of “unglued photography".

Expect: Creative tests (some failed spectacularly) Deep video coherence takes Sound design debates And a suspicious amount of pasta sauce

Bonus tangents: Why product morphs make great ad assets Whether Midjourney understands mood A casual existential spiral about sound design And the weirdest prompts of the week (of course)

It’s experimental AI art meets late-night YouTube rabbit hole. Enter at your own risk.

Every word Read the full transcript