EP 061 · 106 MIN TELL ME WHAT YOU DON'T WANT FIRST.

Live Visual AI AMA: You Asked. And We Went There

January 10, 2026106 min Rory Flynn & Drew Brucker

Watch on YouTube Spotify Apple Podcasts Audio file

The short version

In a live LinkedIn AMA for Episode 61, Drew Brucker and Rory Flynn take audience questions on Midjourney's stalled V8 versus Nano Banana Pro, the difference between Seedream 4 and 4.5, and Freepik's new Variations and camera-rotation tools. The throughline is negative prompting: Rory argues that specifying what you don't want, not centered, not direct-to-camera, not "cinematic perfect", is what keeps images from defaulting to the same bland aesthetic mean.

Key takeaways

01
Kling already ships an edit model; Rory argues Midjourney needs one "yesterday" since it's now table stakes against Nano Banana Pro.
02
Higher base resolution may be why Nano Banana Pro and Seedream nail small text and consistency: more pixels means more fine detail to work with.
03
Freepik and Higgsfield both ran "unlimited" generation promos, then quietly throttled speed or capped them by February, which Rory calls a bait-and-switch.
04
To fight team decision paralysis, build a written visual-identity toolkit of do's and don'ts as an unbiased anchor, and cap revision rounds in advance.
05
A handful of standing test prompts, like a complex mood board or a girl dancing in a mirrored room, quickly reveal whether a new tool is worth adopting.
06
Every model defaults to a bland "mean" look without negative prompts; Rory bans centered framing and cinematic perfection to force a disposable-camera aesthetic.
07
Rory estimates $700-800/month across Weavy, Kling, Suno, and Claude; Drew estimates around $400/month between ChatGPT and Google.

Receipts · said on the record

“Kling has it. Kling has an edit model. Kling didn't have an image model four months ago. They have an edit model. It's like, what are you guys doing?”

Rory Flynn Rory on Midjourney's still-missing edit model versus Kling's

“And then they also started downgrading performance and you had to like pay more money to up the speed. So I was like, yeah, you have unlimited, but you know, now we're going to, we're going to strip the speed.”

Rory Flynn Rory on Freepik and Higgsfield throttling their "unlimited" generation promos

“I'm probably in like a seven eight hundred dollar a month range on this stuff if I had to guess … it is my full-time job. So it's like this is like my whole, like my whole life is this stuff.”

Rory Flynn Rory on his total monthly spend across AI tools

“I think about it like when you talk to your significant other and you're like, what do you want for dinner? And when the answer is so vast, it's like, just tell me what you don't want first.”

Rory Flynn Rory's analogy for why negative prompting beats open-ended prompting

“There's so much value in having been early because you've kind of really put in so many reps to understand how the models interpret things and how just using one word and putting a different synonym in its place can completely change things.”

Drew Brucker Drew on why early prompting experience still pays off

Chapters

Questions this episode answers

Does Midjourney have an edit model?

Not as of this live AMA. Rory points out Kling already shipped one despite launching its image model only four months earlier, and calls an edit model "table stakes" that Midjourney needs immediately.

What's the difference between Seedream 4 and 4.5?

Rory says he prefers Seedream 4, finding 4.5 less obedient to prompts and less consistent with product and character details, even though its aesthetics may be marginally better.

What is negative prompting and why does Rory Flynn think it matters?

It means specifying what you don't want (not centered, not direct-to-camera, not "cinematic perfect") alongside what you do want. Rory calls it one of his biggest predictions for 2026, arguing every model defaults to the same bland, averaged-out look without it.

How much do Drew Brucker and Rory Flynn spend on AI tools per month?

Drew estimates around $400/month between a $200 ChatGPT plan and Google costs; Rory estimates $700-800/month across Weavy, Kling, Suno, and Claude, noting the tools are his full-time job.

What happened with Freepik's and Higgsfield's "unlimited" generation promotions?

Both advertised unlimited access; Freepik quietly added an end-of-February cutoff, and both reportedly throttled generation speed unless users paid more, which Rory and Drew describe as a bait-and-switch.

Show notes

Episode 61 turns the Fast Hours mic over to the audience.

Drew Brucker and Rory Flynn go fully live for an AMA that exposes where visual AI actually stands right now. Not the hype decks, but the messy, useful, (occasionally) frustrating truth.

They break down what Midjourney v8 really signals, why the long-awaited edit model has become table stakes, and how Nano Banana Pro quietly changed everyone’s workflow whether they admit it or not. They debate node-based canvases like Weavy and FreePik Spaces, talk through Kling vs Veo 3 vs Runway for motion, and unpack why so many tools feel powerful yet exhausting at the same time.

Along the way, they tackle... creative paralysis negative prompting resolution myths video realism pricing chaos tool fatigue and the uncomfortable reality that AI creativity is now limited more by decision-making than by capability.

It’s candid and opinionated. And it’s exactly the conversation most AI creatives are already having in their heads.

If you’re using Midjourney, Nano Banana Pro, Weavy, Kling, Veo 3, or just trying to stay sane in the visual AI arms race, this episode is required listening.

⏱️ Midjourney Fast Hour(s) 00:00 – We’re live, welcome to Episode 61 02:45 – What this AMA will really focus on 04:14 – From LinkedIn Lives to a full podcast 05:34 – Midjourney V8 expectations vs reality 08:05 – MJ vs Nano Banana Pro workflows 10:15 – Resolution, text, and why pixels matter 13:26 – Seadream 4 vs 4.5 honest reactions 15:15 – Runway 4.5 and the Nvidia signal 17:59 – Grok as a sleeper visual AI platform 19:42 – Is Midjourney falling behind? 22:29 – Edit models as non-negotiable 24:04 – Node-based tools and FreePik Spaces 28:07 – Camera control and multi-angle tools 31:27 – Tool overload and UX fatigue 36:43 – Creative paralysis and decision overload 41:33 – Gating content, growth tactics, and trust 44:44 – X vs LinkedIn for AI discovery 49:11 – Are LoRAs still relevant? 54:40 – FreePik Variations first impressions 56:08 – How much creators actually spend monthly 01:02:49 – 3D workflows and what’s coming next 01:10:10 – Strategy vs experimentation for teams 01:15:03 – Transitioning from image to video 01:20:21 – Motion capture, Kling, Veo 3 01:22:21 – Has AI killed the creative muse? 01:28:13 – Was learning to prompt a waste of time? 01:31:56 – Dance realism and motion problems 01:34:21 – Where creative AI goes next 01:36:00 – Biggest breakthroughs of 2025 01:39:11 – Negative prompting and visual defaults 01:46:21 – Final thoughts and what’s next

Every word Read the full transcript