When Midjourney is in the room, everyone else is forgotten
Edition #10 - Artificially Intelligent, practical AI for Marketers
Midjourney finally dropped its long-awaited video model. We had been waiting for months!
But it was worth waiting. My X feed is full of cool videos with their model. The Midjourney touch is clear with their cinematic outputs. And it’s a cheap and fast model (unlike Veo3).
That’s bad news for Byetdance, which launched their video model at the same time. Nobody seemed to care, despite it being 33x cheaper than Veo3. But did their model deserve more attention?
It also means marketers may have missed two signals showing where marketing and content creation are heading.
In this week’s roundup:
Midjourney enters the AI video ring
Bytedance and Higgsfield beef up their video capabilities
Canva and n8n show us the future of marketing systems
If you care about staying ahead of the next AI leap, this one’s worth your time.
Midjourney’s long-awaited video model is finally live…
Midjourney has been working on its video model for a while. It’s finally live AND impressive. Yes, Midjourney is in the game.
V1 enables users to animate still images (either generated within Midjourney or uploaded externally) by producing four five-second video clips per prompt. Users can also extend these clips in five-second increments up to a total of 20 seconds (see en example below).
Why it matters:
It’s a cheap and fast model. ~$4 for 20 videos vs. $4 for 1 Veo3 video. Probably a sign that the overall cost of video models will decrease - in such a competitive market, I don’t expect anything else.
Extending a clip duration is the road to more consistency. Consistency is a key challenge for video models (without consistency, many use cases cannot be tackled). It’s great to see Midjourney’s good job on that.
Unlike Veo3, Midjourney’s model is image-only (no audio) and very cinematic. It makes it less relevant for several marketing use cases (e.g., UGC videos). But don’t get me wrong, it remains a very useful model for more “crafted” visuals.
Midjourney is still not made for marketers. The overall product still requires prompt engineering for good results. More importantly, their API is not available yet. That’s sad, because many cool automation-based use cases could be unlocked by the model’s speed and cost.
Going further:
Nick St. Pierre is the go-to person when it comes to following Midjourney news.
Midjourney is not free, but you can watch at cool outputs of people on X:
Example one with complex environments
Example two on anime use cases
Example three is quite stunning
Example four is a great collection of artistic examples
… and it took most people’s attention at the expense of others
Midjourney got most of the coverage because, well, it’s Midjourney. But two other announcements were interesting:
Bytedance launched its video model called Seedance 1.0. Lite version is 33x cheaper than Veo3 and generates good multiple cohesive shots but it didn’t make that much noise. Is there a good reason?
Higgsfield launched its Canvas, a so-called “state-of-the-art image editing model”. Is it that good
Same prompt as above with Seedance - not as good
Why it matters:
Big picture: I repeat myself, but the race to AI video models is good news for everyone. It will drive generation times and prices down, while increasing the overall quality of videos (control, realism, etc.).
Only breakthroughs get attention in such a crowded market. Overall, Seedance is a strong model, but it doesn’t outperform Kling 2.1 (esp. now they have sub-models for more control over speed and cost), Midjourney (esp. the cinematic aesthetics), or Veo3 (which set a new standard with audio + video).
Higgsfield Canvas is not a technological breakthrough but a Flux Kontext wrapper. So yes, it’s very good at placing products into lifestyle images or adding something to an image. And yes it works like a charm with Higgsfield’s proprietary video model. But it’s a capacity that you can access in any other platform integrating Flux (Pletor, Krea…). As often in AI, the value comes from UX and showing people concrete use cases, which Higgsfield did very well.
Going further:
A thread that recaps Seedance outputs. Another good thread like that.
X post announcing Higgsfield Canvas got 1.5M views. “Higgsfield Canvas lets you control everything” and the examples are quite telling.
Two more moves toward the automation of creative assets
I am convinced the world of marketing is changing. The best marketing teams will design systems of agents and control their outputs - almost no more heavy production work by humans.
This week, two interesting pieces of news go in that direction:
Canva acquired MagicBrief, an Australian startup with a simple promise: Launch Ads that win. Concretely, it’s a creative analytics & research tool that tells creative teams what will perform before they produce it.
n8n released its image generator. It’s impossible to be in marketing right now and to have missed the countless n8n influencers pushing their visual production workflows. n8n seems to have noticed the trend.
Why it matters:
If you’re a large company making ads, expect Canva sales reps to reach out soon. This acquisition shows Canva’s increasing focus on enterprise segment, and also their willingness to "power the entire content and marketing workflow" as stated by Cliff Obrecht, Canva's co-founder and COO. Marketing performance is where most of the money is, so it’s not a surprising move.
You should start building AI-based systems early. Even if it starts small. If AI improves 10% every month, the way you produce marketing creatives and content should benefit from that. Don’t only ask yourself “how can I make my current workflows more efficient?”. AI is even more a matter of “What can I do now that wasn’t possible 6 months ago, and that my competitors are not doing yet?”. If this dentist in LA did that, I am sure you can too.
Going further:
I wrote an article about Darbyon, an agency that has understood very well why it’s necessary to build systems.
You may also explore some of the n8n workflows that pop up everyday. To be honest, a lot of them are hard to implement and disappointing (reason why I am building Pletor). But it can already give you a bunch innovative marketing ideas. Check this one or that one for instance.
Another busy week, right? Enjoy a well deserved week-end!