AI Brand Animation Pipeline: From Song to Finished 40-Second Spot

Your brand has a look. Holding that look across twelve AI-generated scenes is where most social teams quit. A character’s sweater changes color between shots, the lighting drifts, a prop jumps to the other side of the room, and the forty-second brand open you pictured turns into a pile of clips that refuse to cut together.

Jeff Sieh built one anyway. On this episode of The AI Social Playbook, he walks through the exact chain behind the animated open for The Makers Table, the Friday craft show he hosts with Katie Fawkes. A song written in Suno. Timing pulled out of Descript as an SRT file. A shot-by-shot storyboard from Claude. Character frames from ChatGPT. The animation rendered in Magnific on Seedance 2.0, then cut to the music in Premiere.

What changes is the size of the job. Jeff finished the piece in one Saturday and estimates roughly $100 in rendering credits for both of the videos he shows. The steps that cost him the most time were not the ones you would expect.

What you’ll take away

Jeff writes the song in Suno before anything else, because the finished audio gives him the timing he cuts the storyboard and the video against.
Character consistency across AI-generated scenes comes from attaching reference images to every single prompt, not from a long-running chat that remembers earlier turns.
Start-and-end frame animation beats text-to-video for branded work because the model interpolates between two images you control instead of reinventing your character on each run.
A cluttered set makes frame mismatch obvious: Jeff’s crowded cabin interior produced visible jumps whenever the two frames disagreed on where a tool was sitting.
Rendering both animated sequences in one pass would cost Jeff around $100 in Magnific credits, so the prompt quality going in decides the bill.
When Claude’s first answer does not work, pushing back on it can produce a better one: Jeff argued his way from a broken browser tool to a custom After Effects rig driven by his caption file.

Timestamp index

00:00 Why Jeff does not call this AI slop
00:37 What this episode builds
02:23 Novelty AI video against a sponsor spot that earns its place
04:19 Guest theme songs in Suno, fifteen minutes each
06:17 The pre-production gap that breaks brand consistency
07:11 The Makers Table claymation open
08:22 The nostalgia argument for sonic branding
09:59 Why the song comes before the storyboard
11:45 Reusing one song for every guest with Suno’s Cover feature
13:26 Talking to Claude through Cowork and Wispr Flow
13:43 Building character reference sheets in ChatGPT
14:57 Claude’s shot list, production notes and video prompts
17:21 Inside the Magnific render dashboard
18:07 What the credits actually cost
20:08 Start frames and end frames explained
22:15 Letting Magnific pick the model
23:07 What this would have taken without AI
25:11 Where the hours you get back should go
26:19 Arguing Claude into building an After Effects rig
28:46 The frame mismatch that wastes renders
29:38 Where to find Jeff

Novelty AI video looks great and sells nothing

Scroll any feed, and you will find someone with a flaming arm. Jeff likes that work, and he says so, but he draws a line between the demo that stops a scroll and the asset that does a job. His test is whether the thing has somewhere to go afterwards.

The example he opens with is a twenty-second commercial for Ecamm, the sponsor of his show Social Media News Live. It is styled as a 1940s USO number, closer to Captain America than to a software ad, and it ends on his affiliate URL. He plays it in the same slot every week, and he reports it has worked well for him.

That framing matters for anyone about to spend an afternoon on this. The pipeline below is not general-purpose video generation. It produces a short, repeatable, branded asset that runs in a fixed slot, which is why the consistency work pays back.

Write the song before you write the shot list

Asked where the biggest failure point sits, Jeff points at the front of the process: “I think they’re not taking enough time in pre-production.” His own pre-production starts with audio.

He writes the theme in Suno, which he pays for and uses weekly. Version 5.5 added a built-in editor, and that solved his original problem. His wife’s favorite version of The Makers Table theme ran one minute twenty-one, far too long for a show open, and before the editor existed, he had to find a segment he could clip and fade. Now he trims inside Suno and generates alternate endings.

Only then does he storyboard. His reason is mechanical: “I always start with the song first,” because the finished audio lets him cut the storyboard and the video to the music rather than stretching the music over a sequence that already exists. He pulls the SRT caption file out of Descript and hands the timing to Claude, so the shot list arrives already aligned to the beats.

One more thing for podcasters. Once a theme exists, Jeff strips the lyrics and generates instrumental versions of the same track for stingers and segues, so the segment transitions carry the same musical identity as the open.

Character sheets, and the rule about attaching them every time

Jeff builds character reference anchors in ChatGPT before he generates a single scene. Katie’s sheet came back with the angles, a color palette, a set of expressions, and details the model picked out on its own, including her fox sweater and a bracelet. He built the same sheet for himself, wearing the shirt he wears on stage.

Then comes the part that contradicts how most people use these tools. He does not lean on the conversation to remember any of it. “Every time I prompt, every time, I had these attached,” he says, and he opens a new ChatGPT chat for each image rather than continuing an existing thread.

Skip that, and the model improvises. Ask a text-to-video tool for a claymation shot of a bearded guy with no reference attached and, in Jeff’s words, “it would just kind of guess every time.” Consistency here is a function of what you re-attach, not what the model retains.

What Claude’s storyboard document actually contains

Jeff works with Claude through Cowork, the desktop app, and talks to it using Wispr Flow rather than typing. The back-and-forth is the point. He tells it when something is wrong, and it revises.

The document Claude produced for the forty-second Makers Table open runs to roughly twelve shots. Each shot carries a start frame prompt, an end frame prompt, and the video prompt to run afterwards. Claude also wrote production notes flagging the details that had to stay fixed across scenes, including which wrist Jeff’s watch sits on, since he is left-handed.

It supplied technical parameters Jeff did not know to ask for. The prompts specify Aardman-style animation at twelve to fifteen frames per second, which is what gives the finished piece its stop-motion cadence. He is straightforward about not having come up with that himself.

Halfway through assembly, he found the sequence needed more room at the front. He went back and asked for an establishing shot while keeping everything else intact, and Claude rewrote the scene and re-timed the surrounding document rather than starting over.

Rendering in Magnific: two frames, one model, real credits

Magnific, formerly Freepik, is where the frames become motion. The dashboard takes a start image, optionally an end image, and the prompt. The constraint to know before you plan a sequence: you can run a start frame on its own, but you cannot run an end frame on its own.

Magnific’s auto setting reads the prompt and the reference images and picks a model. For Jeff, it selected Seedance 2.0. He values the aggregator more than any individual model, because the leaderboard keeps moving. ChatGPT is his pick for image generation today, and he says plainly that could change next week.

Rendering costs money, and the episode does not soften that. “You will burn through credits,” Jeff says, estimating around $100, maybe a little over, to produce both videos in a single pass, depending on how good the prompts were. Some runs get discarded outright. He drops one clip of Katie bouncing with a thumbs up because it was not the take he wanted.

Not everything got fixed, either. In the finished piece, there is a cutout at the top of the frame where he was sitting that he could not prompt away. He left it in, on the grounds that nobody notices it unless he points at it.

Argue with it until it builds you the tool

The most reusable idea in the episode is not a prompt. Jeff wanted an old-school bouncing ball over the lyrics of the Makers Table theme, karaoke style, and had no idea how to build one. He credits Allie K. Miller for the approach he took: argue with your AI, and ask it why it cannot do the thing.

Claude’s first attempt was a browser tool that ingested his SRT file and exposed some sliders. It was wonky. He told it so, in those terms: “That’s not working, Claude.” The second attempt was a rig inside After Effects, an application Jeff already knows as a video producer, with the connector written for him and squash and stretch built into the animation.

That rig saved him hours of keyframing, and he now animates the ball with a couple of keystrokes. His read on it: “Just from arguing with my AI, I figured out a new creative workflow.” The output he kept was not a video. It was a tool he will use on every episode from here.

He describes the same posture as art direction. He tells Cowork what is not landing; it proposes alternatives, and some of those alternatives make the final cut. Having Katie welcome the viewer in the opening sequence came from Claude, not from Jeff, and it stayed in the storyboard.

Where the hours you get back should go

Asked what a social media manager should do with the time this frees up, Jeff answers with client work rather than theory. One of his clients has twelve books. Previously, the work of pulling social content out of them meant reading and highlighting by hand. Now Jeff feeds the books to AI along with what the client has already posted, and generates new content in the client’s established voice. The same client’s books are becoming audiobooks in ElevenLabs, including Spanish in his own voice.

On the question of whether any of this counts as art, Jeff volunteers the answer before anyone asks. He says it is not art, it is him having fun and making something specific for a show he wanted to make. On the slop question, he is firmer: “I don’t consider this slop.” It is personal to him, and it is something he could not have had in his toolkit before.

The old version of this project required studio musicians and animators. That is the comparison worth carrying into a budget conversation.

The gotcha that wastes your rendering credits

One friction point, and it is the one Jeff nominates himself. Your start and end frames have to agree. He was in a hurry on one scene, decided a frame was close enough, and it was not. Move a lamp to the other side of the room between the two frames, and the render produces a visible jump.

Tolerance depends on the shot. A close-up with a soft background will smooth the transition over. His cabin interior, cluttered with tools, punished every mismatch. Read the prompts before you spend the credits, and write the fixed details into them, down to which wrist holds the watch.

Grab The Automated Brand Animation Playbook

The character sheet template, the attach-every-time protocol, Jeff’s Claude storyboarding structure, and the start and end frame checks to run before you spend a credit. Download it here.

Full transcript

Jeff Sieh: Now, I’m not, I’m not saying, like, this is art or anything. Like, it’s not art. It’s me having fun and animating something and doing something specific for a show that I wanna do. But like, I’ve always wanted to have my own theme songs and have songs for guests, like, you know, Kimmel comes in and he has, like, the band playing somebody on.

We can do that now. Like, this is an amazing time for creators, and I know people talk about AI slop, and like, I don’t consider this slop. I think this is very personal to me, and something that I’ve never been able to have in my toolkit that I could do on a Saturday.

Mike Allton: Welcome to the AI Social Playbook, the practitioner’s lab where we break down the exact step-by-step blueprints to help you stop doing more with less and start doing better with AI. I’m your host, Mike Allton. And look, let’s be honest. The modern social media role’s become an exhausting, always-on surveillance operation.

You’re expected to create endless volumes of high-performing video and imagery, keep it strictly on brand, and do it all yesterday. Today we’re skipping the high-level fluff and building the brand animation playbook. My guest is visual marketing expert and video producer Jeff Sieh, and he’s gonna show us the exact multi-tool chain he uses to go from an abstract concept to a fully realized animated brand asset using Claude, ChatGPT, and the cutting-edge Seedance 2.0 animation model in Magnific.

By the way, if you’re listening to us on Apple Podcasts, you can get the full value by heading over to our YouTube channel or the podcast show notes page to watch our screen shares. Plus, we built the exact prompt-level blueprint from this episode as a downloadable guide. Hit the link in the video description to grab your playbook right now.

Hey, Jeff, welcome to the show.

Jeff Sieh: Hey, Mike. It’s so good to be here. I always tease when you when we’re together, you know, you’re one of my BBFs; I always say, before beard friends. So we’ve been friends and hanging out online for a long time now. So thanks for having me on.

Mike Allton: My pleasure. It is scary that I’ve known you that long.

Jeff Sieh: It’s scary.

Mike Allton: And by long, we mean long. That’s fantastic. Long, yes.

Jeff Sieh: That’s right.

Mike Allton: Well, I appreciate you coming on here. I’m truly excited to dive into this, because I’ve personally seen a lot of the images and now videos that you’re creating and have been stunned at how amazing they are. But before we get into those, you told me once that marketers are treating AI video like a novelty rather than a utility.

Why Marketers Treat AI Video as a Novelty

What did you mean by that?

Jeff Sieh: So if you’ve seen … If you’re like me and you see the feed, you see all these people talking and teaching, these gurus showing you how you can have, like, a flaming arm or be a superhero, and I mean, that stuff’s cool. It gets me to stop the feed. I’m a video nerd, a video production guy, so I like to see that stuff.

But I have really been trying to, for my audience, dive into, like, okay, how is this practical? Like, that stuff’s cool, but how can we use it as marketers, as storytellers, as creators for an actual project? And I’ve got some examples I’ll show you. And this all started for my show. I have … The sponsor of my show is Ecamm.

It’s a great software. Love the guys forever. But my show is very retro and kind of ’50s style. And that’s just the way I like it. That’s the way I made it. And I built this commercial that I’m gonna show you for that show, because it has a very practical purpose. It’s my affiliate, and they sponsor the show. So let me play that real quick for you, Mike.

When your show deserves the very best, Ecamm stands above the rest. Ecamm.com/jeff, won’t you see? Broadcasting brilliance easy as can be.

So that’s very Andrews Sisters, right? Like, I really nerded out because one of my first favorite Marvel movies is Captain America: The First Avenger, and there’s, like, a USO show tour on there where he’s having to raise war bonds, and that’s what I was kind of going for: that look and feel and that sound. So I had a very clear idea of when I went in to do it. But it’s, as you can see, it’s very practical. It has my, like, my URL to go, and for the affiliate code, and all that kind of stuff, and that has done really, really well for my show.

I mean, you were just on my show, and that was a section that we play real- it’s 20 seconds, really quick- it’s worked really, really well for me. So that’s what I’m talking about, being very practical with some of the AI video tools that we’re using.

Mike Allton: Yeah, and did you wanna show them the clip that you made for me?

Jeff Sieh: Oh, yeah. So every week I have a guest on my show, usually, and I have been doing theme songs for them for probably the last year, and it’s getting better and better. And so this is the one where Mike was on my show, and we have… I actually animate my YouTube thumbnail, but I also play his song at the same time. So let me go and click on. It’s Creator News Live; AI play is done. Mike Allton is here; he’s number one. You need to build your own digital crew. Mike will show you exactly what to do

Now, all these songs that we’re showing you as examples are created in Suno. Now, that little ditty that I did for Mike, if you notice, it has little keywords in it, as it talked about the digital crew that he does.

He’s the hat number one. So all that stuff I came up with, but AI helped me finesse it into that little song, and I have built a workflow now using Suno, where it only takes me about 15 minutes to create it, because with Suno now, you can actually use a previous one or one that you really like, like audio sound, and build a cover on that.

So my theming continues every week throughout the show so that you can recognize it. It’s still from the same musical library. It all sounds similar, but has different lyrics, but has the same singer because I created a persona that sings the same thing every time. So that’s kind of how we do it for the show.

Mike Allton: And I think it’s that consistency that really caught my eye, and we’re talking about months ago. This wasn’t a recent thing. You’ve been very, very consistent in your visual theming for a long time, and I know, you know, a social media manager, maybe, they’re trying to maintain, you know, strict visual brand standards, yet consistency with some kind of a theme, something like that.

What do you think is, like, the biggest failure point then for them when they try to use the normal generic text-to-video tools?

Where Generic Text-to-Video Breaks Brand Consistency

Jeff Sieh: I think they’re not taking enough time in pre-production. Like, I’ll show you here in a minute how I actually work with Claude, and now I use Claude to actually prompt for the best image model right now, the one in ChatGPT.

I think it has the best consistency. It does the best job. So I actually use, I like Claude’s prompts that I put inside of ChatGPT to run to make those first frames for each of the, the, the images. So let me show you kind of like the breakdown. I do it inside of Suno, but I want to show you first, I think the one that I’m most proud of and my most recent one is I do, the Creator News Live is a Friday show, but then Friday afternoons, I do a carving crafting show with my friend Katie Fawkes, who you’ll see kind of her character in this next piece of animation that we have just put out for The Makers Table, so here we go

Dust and stitches, coffee in a chipped old mug. Plans on a napkin, half-finished things you love. Dust and stitches, it’s craft time again. Two makers, one table. There you go. Two very, very different-sounding songs, but that was all built in Suno. So very, very cool. Suno is one of my… It’s one of my favorite things to do every Friday is to set up for the show, and then I got into what they call the super flow state a couple of Saturdays ago, and just, I cranked out that video that you saw.

It’d been churning around in my brain for a while, but I was able to actually take it and sit down and focus and get that. And that’s about 40 seconds of the Maker’s Table, a kind of intro that I have done, so it was a lot of fun to put together. But it’s so exciting as a creator because we’ve never been able to do this kind of stuff before.

When we were kids, Mike- I mean, I’ll speak for me, but, like, I used to sit and sing commercials, like, driving on road trips, right? Like, we had, like, Kit Kat, you know, had the commercial, Give me a break, give me… I mean, we’d have these Sonic brandings that would happen in our head, you know, and then, like, A-Team, and all the Knight Rider, and all the cool theme songs that would go before our favorite shows.

And sadly, a lot of that’s gone away. I mean, there’s not really any Saturday morning cartoons anymore. Though, I really wanted to kind of bring that back. I think that there’s a whole untapped market in nostalgia, and so I mean, the last, I think, good theme song was Game of Thrones. Like, I mean, they just don’t really do them anymore.

You hit skip on Netflix, and you don’t see them, and I kinda wanted to bring that back for my show.

Mike Allton: I love it, and I, and I love, again, the consistency where obviously you were the character, and then those who don’t know Katie Fawkes, that, that looked just like Katie. Mm-hmm. You guys were consistent through that entire video.

Now, today we’re gonna build a playbook, right, for some of these writing animation sequences. Which one are we using?

Jeff Sieh: So we’re gonna be talking about The Makers Table, because that’s the one that has that claymation style. For those of you guys listening, I’d love for you guys to go look at it and see what I’m, what we’re talking about.

But for me, I thought really hard about, okay, the theme song for The Makers Table is very, very different from the ’50s kind of stuff that I was doing for my show. And I just happened to start thinking about, okay, handcrafted, handcrafted- what does that mean? Oh, claymation has that really handmade look and feel, right? And one of my favorites when I was a kid, or when I even showed it to my kids, was Wallace and Gromit, right? Yeah. If you look at these animations that I did, it’s very Aardman style. Maybe it doesn’t have the biggest mouths, as he always would have with Wallace and Gromit, but it’s that same look and feel.

Why the Song Comes Before the Storyboard

And I was able to take that idea and take it into Claude to actually get me that idea. Now, what I usually do, and the way my workflow works, is I always start with the song first. Like, I think of an idea. ‘Because that way I can cut the storyboard, and I can cut the video to the music. That’s just the way I work.

It’s easier for me to fit it in that way. So I actually started, and I’m gonna do a screenshot here of Suno. And I pay for Suno; it’s a paid program, because I use it every week. And I make numerous versions of this. I had a really great one that my wife loved of The Makers Table, and the problem was it was like a minute and 21 seconds, and that was one of the big issues when creating stuff at the beginning.

And so I … The cool thing now with the advancements with Suno is it even has a built-in editor, and I can, like, trim it and then give myself different versions of it. Ooh. And different endings. And so I had … The big issue I had before was trying to create these short versions for my show or commercial-kind of lengths, and now you can do that inside of Suno, especially with the new version 5.5.

You also have what’s called Voice up here. You can actually sing to Suno, and it can do it in your voice. A little scary; I don’t use mine. But I have some already. Like, I have this one: Makers Table country, male. That’s the sound, the guy that you heard. And then I also, for my show, I have this female scat lounge singer, ’cause she kinda has some harmonies that she goes in there.

But you can create your own persona, or your … or find one you like and then save it, and then you can create whatever you want later based on those versions. And then the other cool thing is, once you get, like, say, a song that you really like, you can do different versions of it. This one’s, I think, an ’80s one.‘Cause it’s craft time again to make us one table and room for just silly stuff, right? But let’s say I like that one. I can just go over here to Remix and go to Cover. And it takes that exact song, and then I can change the lyrics up over here and fit it for, like, if a new guest is coming on, like we did that example of Mike or whoever.

You can actually keep that same theme and style and have new lyrics on top of it. So you have that consistency throughout your show or throughout your commercials or your campaign or whatever you need to do by doing it that way inside of Suno.

Mike Allton: Love it. That’s fantastic. And now I need to go back to Suno ’cause- right, the last time I used it was probably a year ago or so. Yeah. I used it as the theme song for my own AI podcast, and I loved it, but it didn’t have that kind of editing capability back then. So really it was me; it was more about me finding the perfect segment to then clip it off and, right… You know, fade out or something like that.

I didn’t have that kind of fine-grain control. So step one is: if we’re doing audio, if we’re having music, use a tool like Suno, right? Right. But then you’ve got a storyboard. You’ve talked about that a little bit more. Mm-hmm. And I’d love it if you’d just elaborate for us. You talked about chaining a bunch of tools together, starting with Claude.

Walk us through how you’re prompting Claude to first lay out that sequence – and how you ensure that it actually outputs, you know, functional image prompts rather than just, say, narrative stories.

Jeff Sieh: Yeah. So real quick, back, back to Suno, just for the podcasters, the audio people out there, once you get a theme down, you don’t have to use lyrics.

So I have that same theme for my show, but I have no lyrics for it, so I have opening music. And I’ve actually been doing this- right… for a lot of my podcaster friends. It’s like if they need segues or stingers or things like that in the same theme, it’s really great now that you can do all that stuff inside of Suno.

So I just wanted to say that really quick. Let me show you kind of what Claude produced for me. Now, I want you to… This is the end result. You gotta think in your head, like I use Cowork for this- my desktop version of Claude. I have used Cowork. I just like the way that I can talk back and forth with it.

Building Character Reference Sheets in ChatGPT

And so sometimes I would use the start and ending frames; sometimes I’d run the video prompt with just the first frame and the way that looked inside of Magnific. But here you see all the different animations that I’ve run for each one of those scenes. And the cool thing is, you can go up here, and you can go into Recreate, and it actually will keep the image, and this is where you drag the start image and the end image.

This time I just used a general image because I was experimenting with some different things, but it’s very, very cool if you need to run it. Now, you will burn through some credits. Like, these are paid programs. Suno’s paid. The Magnific is paid. You could actually not use ChatGPT and do it all inside of Magnific because they have access to those image generation tools as well.

But I already had ChatGPT, and you will burn through credits. So I would think for both of these videos, if I wanted to do them all in a single time, it’d probably cost me around $100, maybe a little over, depending on how good my prompts were. But you can see here, this one just went off the start image.

I did not use this clip ’cause she’s kinda jumping up and down with her hands, you know, giving me a thumbs up. But you can see I’ve worked through some of these different scenes to get exactly what I wanted. Once these were all done, I just dropped them into an editor. I use Premiere, and I just timed it to the music, and that’s how easy it was.

I did do some multiple shots of the logo re- kind of resolving at the end, and if you watch that video again, you’ll notice at the top, there’s a cutout where I was sitting, and I could not get it to not prompt. But nobody notices it unless I point it out to them… so I just kinda ran with it. But that’s kind of in a nutshell how, you know, how we did all this quickly.

Mike Allton: Awesome. I wanna see more Magnific in a moment, but let me go back- okay… to a couple things that, that stood out to me. First, you talked about how important it was to have not only that, that graphic that illustrated all the different scenes, everything like that. Mm-hmm. But I’m pretty sure I heard you say every single time you were prompting Claude, you were reattaching that image. Did I hear that right?

Jeff Sieh: Yeah. So Claude, you know, kind of at the beginning, but for ChatGPT, because remember, all these prompts are for ChatGPT, because it has the best image generator out there at the current time, in my opinion. So I would attach those reference image, im- images every time with that Claude prompt. So I would take those reference images of the Aardman characters or the claymation characters and attach those every time I went to generate that beginning and ending image prompt that I would use in Magnific with that, you know, kind of reference style of Katie and I’s character. So I always had those characters attached to whatever prompt I ran, so everything would stay consistent.

What Start and End Frames Actually Do

Mike Allton: Got it. And could you talk to us more about- you mentioned start frames, end frames. What is that really? So, you know, someone watching this or listening to this right now, maybe they haven’t really worked with video in that way. What are those, and why are they so important?

Jeff Sieh: So you can do, inside of any of these animation kind of tools, these AI animation tools, you could just do a text-to-video prompt.

Like, I would just go in there and say, “Hey, I want a claymation-style thing of a bearded guy,” whatever, and it wouldn’t get my character. It would try to generate what it wanted, and if you didn’t have a reference image for those characters, it would, it would be all over the place. Like, it would just kind of guess every time.

So having very specific reference images for each prompt is what makes it happen. Everything I went to had a reference image inside of it, and so I would make sure every time that I would attach lots of times I would attach the Claude starting image, and I would also attach the reference images of Katie and myself as those characters.

Mike Allton: Got it. Got it. So now we’re looking at Magnific. Go ahead and show us a little bit more about this rendering dashboard.

Jeff Sieh: Okay, so one of the cool things I like about Magnific, now it used to be called Freepik; Magnific just bought them not very long ago, is right here, every time, the start image, you can actually drag whatever you want from your computer.

These are some that I’ve already had inside of here. Or you could take a photo, whatever you want. But you can’t have just an end image. You have to either have a start image or a start and end image. You can’t just start with an end image. It just doesn’t work that way. So all of mine at least have that reference image.

Here is the start image, and then right here is the actual shot. And remember, with that Claude workflow that I showed you, it already gave it to me, right? It already gave me the prompt to use for this animation in the storyboard. So this… And sometimes I would go in and change it if it was doing something wonky.

Like, I changed this a little bit because it wasn’t getting the come here gesture enough. It wasn’t big enough, so I would come in here and actually change it. A lot of times I wouldn’t. I would just use whatever that Claude would have. Like this one, it was too slow, so I would come up and I did a little bit differently.

I didn’t like how the mug went up, so I changed that. So this is where you can go in and edit it. The cool thing about Magnific is you can actually do auto. It will look at your prompt and your reference images and pick the best model for you. And so when I started, it went to Seedance, which is kind of the one everyone’s talking about right now.

But using these aggregators, these tools that kind of can use all the different models, is really handy because, as I said, right now, ChatGPT is the best, I think, image-generated model. But that could change next week. Like, really, it could. Like, they change so fast. And having a place where you can go in and pick that model, or they add that model really, really quickly- Dustin from Magai does the same thing. Having that ability to test the latest model is what kind of gets you ahead and saves you a lot of heartbreak later on when you’re spending credits.

What This Would Have Cost Before AI

Mike Allton: So if you’d have been doing this without AI’s help, how long would it have taken you?

Jeff Sieh: I’ve always wanted to do stuff like this, right? Yeah. Like, we’ve all wanted to do really cool opening animations for whatever project we’re working on, at least if you’re a nerd like I am.

I just really think that I wouldn’t have been able to; this is a dream. I would have never been able to do that, ’cause you would have to hire studio musicians. You would have to hire animators. You could maybe sketch out something and do it, but now you can take what’s in your brain, mull on it for a couple days, kinda think it through, and then create something that, like, wow.

Like, you’ve never been able to do this before. Now, I’m not, I’m not saying, like, this is art or anything. Like, it’s not art. It’s me having fun and animating something and doing something specific for a show that I wanna do. But like I’ve always wanted to have my own theme songs and have songs for guests.

Like, you know, Kimmel comes in, and he has, like, uh, you know, all the band playing, somebody on. We can do that now. Like, this is- yeah… an amazing time for creators, and I know people talk about AI slop, and like, I don’t consider this slop. I think this is very personal to me, and something that I’ve never been able to have in my toolkit that I could do on a Saturday.

Like, that blows my mind

Mike Allton: Yeah, no, this is definitely not AI slop. To your point, this is business; this is marketing materials. Yeah … this is content, a new form, a new way of creating content, something that, to your point, if you would’ve even done it before, it would’ve taken you multiple days with a big budget.

Now it’s possible with, you know, very, very little budget, very little time, which means it’s, it’s an option now for people for whom the previous way just wasn’t an option. So you’ve got, you know, maybe a social media manager who’s listening, who’ve figured this out; they’ve automated this part of the robot work.

What kind of high-level strategic initiative should they be reallocating some of that saved time towards to finally prove their value, do you think, to executive leadership? But this is something that they were already spending a lot of time doing, and now they’ve got AI helping them.

Jeff Sieh: So for me, I think it, it lets… And, and a lot of the clients that I work with who consult with me about this kind of stuff, is like I’m freeing them up now to, to have more creative control about other things that they haven’t been able to do before. Like, I’m working with a guy who’s got, like, 12 books, and now he can take this content from his own books and create things that he can put out on social media that he hasn’t been able to do before.

He’d have to go manually through and, like, highlight things inside of it. But I can now take AI and have him look at it and, like, okay, this is what he’s posted before; this is his voice. What can we take out of that and create something new very, very quickly? That’s never happened before. Like, another example is, like, he’s got all these books that he’s, he wants to make audiobooks.

I can now do that in ElevenLabs. I can now do that in Spanish in his voice using, you know, that stuff. So I mean, like, those kinds of things I think are just amazing. You just have to sit there and think about it and what you can do. For me, like, here’s the next step for this. Because that song is so very catchy and earwormy, I wanted to have, for the Makers Table, when we play it on our show opener, a bouncing ball- the old-school kind of animation that, oh, yeah

Arguing Claude Into a Custom After Effects Rig

Like karaoke, but I didn’t know how to do it. So I was there, and I just asked Claude. One of the m- the best things that I’ve ever heard is this is from Allie K. Miller; she goes, “Just argue with it. Like, argue with your AI and say, ‘Why can’t you do this?'” Right? And so I did that, and it built me this online tool that took my SRT files, which are caption files, and I had these sliders, and I could adjust them.

It was still wonky, and I said, “That’s not working, Claude.” And it said, “Okay, let me build you a rig inside of After Effects,” which I, as a video producer, I know how to use. But it saved me hours and hours of keyframing stuff, and it has the squash and stretch and all the cool stuff that you hand animators can do.

It built it, it built it. It did a connector inside of After Effects, and I can just animate it by tapping a couple of keys now. And I’m like, “Holy cow.” Like, just from arguing with my AI, I figured out a new creative workflow that I can use. So that’s the kind of stuff I think that creators need to, and social media managers and people in the advertising space are starting to argue.

It’s like, “Hey, I’ve not been able to do this before. Can you do it now?” And see what it says. Most of the time, it’s gonna create or start walking you through things. I mean, that’s what’s so amazing to me about Cowork is like, “Hey, I can’t do it this way, but maybe if we tried this way, we can do it.” And you have this, like, collaboration.

It’s like I had a whole team of artists that I could art direct, that I could, like, “Mm, I don’t know what I’m trying to say here, but this is what…” And then it would suggest things. I’m like, “Oh, yeah, let’s go that way. Let’s try having, you know, the cabin exterior shot with smoke coming out of it.” And like it said, “Oh, you welcome Katie at the beginning.

Let’s have Katie welcome the viewer in that sequence.” And I’m like, “That’s a great idea. Let’s do that.” And so that was added; that came into the process later on down the line as we kind of storyboarded it out. So that’s the kind of collaboration, thinking of it as an art director or s- an artist that you can direct, I think is the kind of mental unlock that I had to do.

Mike Allton: That’s huge, and I also think it’s helpful for those of you listening who maybe have questions, you’re not sure where to even start, you can go to AI for that as well. For, “Hey, how do I do this with you? How do we get started? What’s an example of a prompt?” That sort of thing. Mm-hmm. Of course, we’re gonna give you all the prompts, but, you know, those of them you guys are watching, you wanna build this exact pipeline today.

What do you think, Jeff, is the one kind of gotcha or hidden friction point in the Seedance model that they need to watch out for so they don’t waste all their rendering credits?

The Frame-Consistency Gotcha That Burns Credits

Having a start and end frame is key, but sometimes even with ChatGPT, you’ve really got to look at the details. I was in a hurry, and I was like, “Oh, that was close enough.”

But even if a lamp moves to the other side or something, you’ll have a weird jump. And so really look and force that AI to have a consistent start and ending frame. Now, if it’s a close-up of somebody and it’s like, you know, there’s kind of a blurry background, it’s not gonna matter as much because it’ll kind of smooth that transition over.

But my, that cabin you saw was very cluttered with tools and stuff, and if it changed too much, it was a big jump. So really be consistent on your, on setting up your character sheet and setting up those prompts. Read through them and make sure that, you know, “Oh, hey, by the way, I’m left-handed. I’m always gonna have the, the watch on this wrist.”

So make sure you put those types of details inside of your prompt and keep, keep it consistent as it goes down through the storyboarding.

Mike Allton: Jeff, this has been just an absolute masterclass in tactical execution. Folks, if you want the exact prompt streams, the custom Claude storyboarding templates, and the step-by-step screenshots of Jeff’s Magnific workspace, don’t try to scribble it all down.

We’ve organized it into a single downloadable asset. Click on the link right now in the description to grab the automated brand animation playbook and put it to work today. Jeff, for folks who want to know more about you, where can they go to reach out?

Jeff Sieh: I am @jeffsieh everywhere. That’s J-E-F-F-S as in Sam, I-E-H.

I before E, especially in C. That’s how my mom taught me to say it. So yeah, I’d love to connect wherever you guys are.

Mike Allton: Fantastic. Jeff, thank you so much for setting the gold standard, I think, for what a real playbook looks like. Until next time, stop feeding the beast and start leading the machine. That’s how real practitioners are winning with AI on social.

The AI Social Playbook is brought to you by Agorapulse and archie.app. Make sure to subscribe to the channel on YouTube so you don’t miss a single hands-on session, and grab the link in the description for the complete episode blueprint. See you in the lab.