I built a text-to-video AI in 30 days.

Here is the link if you want to roast the LP: textideo.com

This is such a real post. I’m currently building Matcha Resume (AI resume optimizer) using Next.js + Supabase as well, so I feel that "AI Tax" pain deeply.

To answer your pricing question: Do not do unlimited flat-rate subscriptions. Since your COGS (Cost of Goods Sold) varies per user usage (GPUs/APIs), an unlimited tier is a death trap for an indie.

I’d recommend a Hybrid Model (Subscription + Credits):

$X/month: Includes Y videos (credits).

Top-ups: Pay-as-you-go for extra credits.

This secures your MRR (predictability) while protecting your margins from power users who would otherwise burn your API budget. It aligns your revenue with your costs.

Also, the pivot from "Generic Video Tool" to "Newsletter-to-Video" is smart. Niche workflows > GenAI wrappers.

Keep shipping! — Abhi

Really cool breakdown of your journey thanks for sharing it so openly. Your pivot makes total sense too; most tools only take off once the creator stops building and starts talking to real users. I’ve been through something similar when I built a net salary calculator and realized the niche mattered way more than the tech.

Your pricing struggle is real for API-heavy products. A lot of indie founders I know start with a credit system first to protect margins, then add a light monthly plan once usage patterns are clearer. Might be worth testing that combo.

@Smoteria

Love the clarity in your positioning shift. The 10x lift makes total sense,platforms reward specialization.

Funny enough, this is the exact pattern I see when helping SaaS founders grow through Reddit: the moment their message becomes hyper-specific, subreddits respond insanely well.

If you ever want to test how niche-focused Reddit marketing could amplify your SaaS design presence, happy to share a quick breakdown.

Man, reading this felt like looking in a mirror from 6 months ago.

I had the exact same “content treadmill” pain — folders full of scripts that never became Shorts because editing took 4+ hours and every tool wanted $50–$90/month. I was literally down to my last R400 and almost gave up.

So I did the only thing I could afford: built my own version in 72 hours straight.

AI Video Narrator is basically the “I’m broke and desperate” edition of what you’re building — pure client-side, no backend workers, no GPU bills, just Gemini + Web APIs. Takes a script → narrated + captioned Short in ~45–60 seconds.

I’m keeping it stupid-simple on purpose because I never want another creator to have to choose between rent money and video tools.

Just launched yesterday with a $45 lifetime deal for the first week (unlimited generations, better voices) because subscriptions were the whole reason I had to build this in the first place.

Respect for shipping and being brutally open about the bills. That’s the kind of transparency this community needs more of.

“nobody searches for a general tool” nice idea to keep in mind always it is easy to get lost in trying to make something work. On pricing for API-heavy products I’ve seen hybrid models work best. a small monthly base for commitment and predictable revenue plus credits for heavier usage. The logic is it protects margins without scaring off light users and big companies do it like this too. Thanks for your experience sharing!

Hey! Cool idea. I'm actually looking for the opposite - I need a growth partner for
my existing SaaS.

If you know any growth-focused people looking for equity opportunities, send them my
way!

Good luck with your search.

I have similar struggle. I created Container Load Optimization Solution available as SAAS.

1. I entered a highly competitive market where free tools are already available.
2. Porting users to paid tool is very much hard.
3. I do have monthly fixed cost to cover.

I have visited your site and text to video section show case beautiful videos, My advice to you shall be: Add the text along side to generated video. Users shall be available to visualize the input to output. Show them the value.

The feedback cycle is too long. So be patient on DMing.

Your customers are on Social media, You may think of automating the process, for example people submit the text and you post it on their social media (or send them interface to publish). The end to end integration can be your differentiator.

The "Text to Video" is too generic.

Consider the following benefit driven approach (I learned from Indie Hackers) :
Before:
We maximize container space utilization and significantly reduce your freight Cost.
After:
You will never see a partially filled container leave your factory gates again, maximizing profitability on every shipment, resulting in Zero dead freight and Shut-outs.

You may advice me on my pricing strategy: https://www.loadviewer.com/introduction/plans

Thanks

That’s an impressive achievement — building a text-to-video AI in just 30 days shows real drive and technical skill. If you’re exploring data-driven or AI-powered tools, AltPaths offers powerful crypto analytics and market screening to help support smarter investment decisions.

Simplicity is key for creators. Focusing on a clean, distraction-free editor is absolutely the right vision.

You could do a combination of monthly sub and a credit system. Just make sure it is extremely easy to understand the tiers. E.g.

• Free: 1 video per month
• Basic: 29.99 per month (cap at 10 videos, use credits after that)
• etc.

“Wow, that’s impressive! Building a text-to-video AI in just 30 days shows incredible focus and skill. Excited to see the results and how it handles creative, real-world use cases.”

Really appreciate how openly you shared the process — this is the side of AI tooling most people don’t talk about. Your stack makes sense, and the cost breakdown is a good reminder that “AI SaaS” isn’t anywhere near traditional SaaS margins. The pivot toward a specific workflow niche was smart. Content creators don’t buy generic tools; they buy shortcuts for their exact pain.

For pricing, most API-heavy products I’ve seen end up using a hybrid model: a small monthly base + credits. The base covers fixed server costs, and credits protect you from power users eating your margins. It also keeps casual users from feeling locked into high plans.

You’re on the right track. Getting your first paid user through manual demos validates that people will pay once they feel the workflow solving something real. Keep iterating on that.

impressive and cool project

Really appreciate you breaking this down so openly. Honestly, it’s wild that you got all that done in a month, the product looks super polished already.

I’m also exploring AI-driven tools and completely get the pricing dilemma. Credits make sense from a cost side, but users really don’t like thinking in “tokens.” It’s a tough balance.

I’m curious where your backend is hosted, and why you still needed FFmpeg if you’re using newer video models like Veo or Sora. Seems like a smart choice for full control, though.

Also wondering how you plan to stand out with so many AI video tools popping up every week. Still, huge respect for what you’ve built. Hope the next launch goes a lot smoother.

I visited your website, looks similar to leonardo ai. It would be nice to show the prompt that's been used to generate the sample video. Plus, I wasn't able to generate something, not even for once without signup. And how are you using the AI? Hosting the models yourself or relying on API providers?

Just commenting to say I admire your commitment. It takes a lot of determination to keep going when you get a bumpy start. Glad it's paying off.

The amount of convenience we provide is based on our infrastructure. Initially, it is confusing to price. So,rethink your products value and manually do the work of an PRO guy in your niche. The result a content maker & producer will depend on their tools & workflows in their workforce. So check the usage of those tools & $$ they provide as salaries.And do the thinking of how much results(shorts,videos,reels,etc) they produce per month and make a calculation of productive time consumed by their workforce(employees). So if you offload the mental load of the user & speed up the workflow between the workforce or replaces a percent of them to produce results efficiently you can be confident in pricing. Be delusional or DO the manual work and Ask yourself "How much would you pay?". And Ask or discuss with your users itself or with your AI personas "How much would they pay? whether monthly or pay as they go or yearly? ". Atlast, observe everything around which you build and invest accordingly and price. If you completely observe it,you can price it confidently and change by features and users.Price for the user first,they would love it if the price is affordable from their perspective and later price for yourself and the company bu doing long term calculations. So set an average and then raise accordingly and respectively.

As a matter of fact, it is better to have payments fast and monthly subscription should not be mandate. People will pay and use just as per convenience. You don't use 2.5 GB/ DAY everyday as a normal person who purchase a monthly data pack.

ohhh nice

Wow, really impressive work building a text-to-video AI in just 30 days! 🎉 Your breakdown of the stack, server costs, and the launch flop is super transparent and valuable for any indie hacker.

I totally relate to the “content treadmill” problem—spending hours on small edits feels like grinding in shadowfightmodapkcom: every little move counts, and if you don’t optimize, you burn out fast. I love how you pivoted to niche outreach and manual demos—that’s exactly the kind of strategy that wins when the big players dominate the generic space.

I'm currently learning AI and building some tools and websites with Cursor; I hope to launch my first site soon, just like you did.

Base monthly plan + usage based credits.
Something like:
A small monthly fee that covers your fixed costs
A set number of video credits included
If they need more, they buy extra packs
That way:
Casual users stay happy
Power users pay for what they consume
You protect your margins
You still get predictable MRR

"I Built a Text-to-Video AI in 30 Days" is an inspiring account of rapid AI development, detailing the challenges and breakthroughs in creating a system that converts text into video. The author shares practical insights, tools, and lessons learned, making it both motivational and educational for AI enthusiasts and developers interested in generative media.

Love seeing honest build-in-public stories like this.
I’ve been around founders for 15+ years and the pattern is always the same — things work the moment you stop pretending and start telling the truth about what’s actually happening. Respect for sharing the real journey, not the polished one. Following along.

Building an MVP in 30 days with clear constraints is admirable discipline. Your transparency about costs is refreshing - the AI tax is real and catches a lot of builders off guard. The pivot from self-serve to manual demos shows good market instinct. Getting that first paying customer through direct outreach is often the hardest part.

Wow, that’s impressive! Building a text-to-video AI in just 30 days shows remarkable focus, technical skill, and creativity. It’s exciting to see rapid innovation in AI, turning ideas into visual content so quickly. Curious to learn about your workflow, the challenges you faced, and how you handled video generation and optimization efficiently.

That's impressive, man.

good luck :) I just built one , image to video for AI UGC videos

This is really cool and you have a fascinating tech. Definitely want to learn more not just in your tech, but in how your solution is differentiated from direct AI video generating promts

The main takeaway is that AI tech is amazing, but cloud costs and positioning are critical. Video rendering, API usage, and GPUs add up fast, so a generic "text-to-video" approach often flops without a clear target audience.

This is really cool, I worked on a similar project https://vidbuilder.ai but not using Video models its for Text and Animation based videos. I would like to hear your comments and maybe we can join forces together

That’s incredible! Building a text-to-video AI in just 30 days shows remarkable skill, focus, and creativity. The potential applications—from content creation to education—are huge. Excited to see how it performs and evolves. Truly inspiring work for anyone interested in AI and multimedia innovation.

Love the "stopped coding and started DMing" lesson. That's such a hard shift to make as a developer—we want to believe the product will speak for itself.
On the pricing question: I'd suggest starting with a credit system instead of subscriptions. Here's why—users are hesitant to commit to monthly fees for a new product they haven't tested thoroughly. Credits let them try it with low commitment, and you can always migrate to subscriptions once you have a solid user base and proven retention.
The transition path could be: credits → credits + subscription bundles → pure subscription tiers. Lets you validate willingness to pay without scaring away early adopters.

Damn, this is one of the most honest breakdowns I’ve seen about building an AI video tool — especially the part about the “AI tax” nuking your margins. The pivot to doing manual demos was a smart move; most people never talk to their actual users. Curious to see where you take it next, especially how you tackle pricing without letting power-users melt your GPU bill.

Wow, building a text-to-video AI in just 30 days is incredible! This showcases impressive technical skill, creativity, and dedication. The potential applications—from content creation to education—are huge. Excited to see how it performs and evolves. Truly inspiring work for anyone interested in AI and multimedia innovation.

This is super impressive 30 days is a very tight build cycle for anything video-related, especially when dealing with rendering + model latencies.

I’m curious about two things from your journey:
1. Which part took the most time to get right?
The model workflow, video pacing, or getting consistent visual quality?
2. How did you validate that creators actually wanted this format?
I’m building something for creator workflows myself, and picking the right use case has been harder than the technical side.

Really admire the speed and the dedication. Would love to hear more about what surprised you the most during the build.

Loved the level of detail here.
The job queue rewrite and async rendering were definitely the right move — cloud video assembly is brutal without proper workers. Also agree that stability beats model choice 100% of the time.
On the pricing side, most AI tools I’ve built ended up using a “base subscription + metered usage” model. It filters out abusers while keeping normal creators happy.
Thanks for sharing the honest numbers and lessons — super refreshing to read.

Speed of execution is everything here. Great job on shipping this fast! 🚀
With giants like Sora and Runway dominating the general space, do you plan to niche down to a specific vertical (e.g., marketing, faceless channels, or education)? I feel like focusing on a specific use-case is the indie superpower right now.