We Built an AI Video Tool Because Hiring Animators Was Too Expensive (And Existing AI Tools Frustrated Us)

By Seagnson

June 4, 2026

## This is not a "we grew to 10k users in a week" post

Because we didn't.

We launched SeedClip exactly 4 days ago.

Right now, we have 127 users. Most of them are friends, a few are strangers who found us on Reddit.

I'm writing this on Indie Hackers not because we've made it, but because I want to document the journey from day one.

And maybe — if you're building something in the AI space — you'll find this useful.

---

## How it started: a birthday gift gone wrong

My girlfriend's birthday was coming up. I wanted to make her something personal.

She loves fantasy webcomics. I thought: what if I make a 30-second animated short featuring her as the hero?

Simple, right?

Wrong.

I looked into hiring animators. Too expensive. Way outside my budget for a personal project.

So I turned to AI video tools.

---

## The AI tool rabbit hole (and why I got stuck)

I tried everything.

Runway Gen-2. Beautiful results. Slow renders. Expensive if you iterate a lot.

Pika Labs. Fun and fast. But character consistency? What character consistency? Every clip looked like a different person.

Kling. Impressive motion. But the queue times were brutal. I waited 15 minutes for a 5-second clip.

AnimateDiff. Powerful but complicated. I spent three hours installing things and gave up.

Here's the pattern I noticed:

Every tool was great at one thing. None of them were great at making a coherent short story.

I wanted:

- The same character to look the same across multiple clips

- Fast iteration (I don't want to wait 10 minutes to see if my prompt worked)

- Affordable pricing for a hobbyist

I couldn't find it.

So I decided to build it.

---

## The stupid decision

I am a developer. Not a great one. A decent one.

I know how to call APIs. I know how to glue things together. I do not know how to train diffusion models.

But I figured: how hard can it be to wrap a few existing models into a simple interface?

That was naive.

---

## What actually happened

I spent 20 days building a prototype. Alone. At night. After my full-time job.

Day 6: I built a simple UI that called multiple APIs. You could type a prompt, pick a model, and generate a clip. It worked. Barely.

Day 14: I tried every model I could find. Runway. Pika. Kling. Some open-source ones. They were all fine. But nothing solved the character consistency problem. Faces kept drifting between clips.

Then I came across ByteDance's Seedance 2.0.

I ran a few tests. Same prompt. Same character reference. Different clips.

The results were noticeably better. Faces stayed consistent. Backgrounds didn't warp. It wasn't perfect, but it was the closest I had seen.

So I stopped trying to build my own model.

Instead, I built a clean, simple interface around Seedance 2.0. A website that doesn't hurt your eyes. A workflow that doesn't make you want to quit.

That's it. I didn't reinvent the wheel. I just put a better handle on it.

I called the website SeedClip.

---

## The launch (humble numbers)

We launched on a Thursday night.

I luanch on Product Hunt, no posted on one subreddit r/aivideo. No Twitter thread. Just a simple post:

> "I built an AI video tool that tries to keep characters consistent. Free tier. Would love feedback."

Day 1: 23 users. Two of them were me testing with different email addresses.

Day 2: 41 users. Someone on Reddit said "this is actually not terrible." I screenshotted it.

Day 3: 67 users. The first stranger generated their first clip. I watched the log and felt something I can't describe. Pride? Relief? Both.

Day 4 (today): 127 users. 340 clips generated.

---

## What Seedance 2.0 & Seedance 2.5 actually does (no hype)

Let me be clear: we are not competing with Runway or Kling. Those are amazing tools built by real AI researchers.

We are solving a smaller, more specific problem.

Character locking. You upload 3-9 reference images of your character. Seedance 2.0 tries to keep that character's face, outfit, and silhouette consistent across multiple clips.

That's it. That's the main feature.

Other things we focus on:

- Speed: Most clips render in under 60 seconds. Not because we're geniuses. Because we optimized for speed over absolute quality.

- Affordability: Free tier gives you 30 credits. Paid tier is $15/month. We're not trying to get rich. We're trying to cover server costs and keep building.

---

## The flaws (I'm not hiding them)

Here's what doesn't work well:

Complex motion breaks. If you ask for a character doing a backflip while catching a sword, limbs will disappear.

Not great for realism. Our model works best for illustrated/anime styles. Realistic human faces? Sometimes weird.

The UI is ugly. I am not a designer. It looks like a developer made it. Because a developer did.

Bugs. Lots of them. We fix one, two more appear. That's just where we are.

---

## The business model (or lack thereof)

Right now: $0 MRR.

We have a Stripe account. We have a $15/month plan. No one has bought it yet. And that's fine.

We want people to use the free tier until they hit limits that genuinely frustrate them. That's when we'll know we've built something worth paying for.

Goal for month 2: get to 20 paying users. That's $180 MRR. Enough to cover server costs and buy me a few sanity coffees.

---

## Why I'm posting this on Indie Hackers

Three reasons.

1. Accountability. If I post my numbers (127 users, $0 revenue), I can't pretend things are better than they are. It keeps me honest.

2. Feedback. We're early. Like, really early. I want to know what features actually matter to people who make videos.

3. Documentation. There are a thousand "we grew to 100k users" posts. There are almost no "we launched 4 days ago and here's what happened" posts. I want to be one of the few who shares the boring, messy beginning.

---

## What's next

This week:

- Fix the top 5 bugs from user reports

- Add a simple tutorial video (because the UI is confusing)

- Respond to every single support email within 2 hours

This month:

- Batch rendering (queue up multiple clips)

- Better character fine-tuning (fewer melted faces)

- Proper documentation

If we ever hit 1,000 users:

- I'll write part 2 of this post