Why I Built MP4Transcript — A Smarter Way to Convert Video & Audio to Text

dragong — Thu, 06 Aug 2026 15:50:48 +0000

Every side project starts with a problem you can't ignore. Mine began with a 90-minute interview recording and a deadline.

The Problem That Started It All

I needed to transcribe an interview for an article. The audio was clear, the speaker was articulate, and the content was exactly what I needed. But transcribing it by hand? That meant 90 minutes of listening, pausing, typing, and inevitably missing something important.

I tried the usual solutions. Most were expensive. Some required uploading my file to a server I didn't trust. Others gave me timestamps that were off by seconds, or exports that didn't work in any tool I actually used. And nearly all of them wanted a monthly subscription — even if I only needed to transcribe something once a month.

That's when I realized: the market was solving transcription for enterprises, not for individuals who just need to turn an MP4 into text occasionally.

So I built MP4 to Transcript.

What MP4Transcript Actually Does

MP4Transcript converts uploaded video and audio files into editable, timestamped transcripts. It supports:

Video: MP4, MOV
Audio: MP3, M4A, WAV

Once transcribed, you can export in three formats:

Format	Use Case
TXT	Clean text for notes, research, drafts, archives
SRT	Timed subtitles for video players
VTT	WebVTT format for web players and captions

The workflow is simple: upload → review → export. No complex settings, no confusing options.

What Makes It Different

1. No Subscription Required

This was non-negotiable. I didn't want another monthly bill for something I might use twice a week. So MP4Transcript uses one-time minute packs. Buy credits when you need them, use them when you have files. No commitment.

2. Credits Are Protected

If a transcription fails, you get your credits back. If your file is shorter than the initial estimate (say, you estimated 30 minutes but it was only 25), the unused credits are returned automatically. You never lose money on a failed attempt.

3. Privacy by Design

Files are stored privately in Cloudflare R2 and automatically deleted after 72 hours. No public listings, no sharing your transcriptions with strangers. You can also manually delete a transcription anytime.

4. What You See Is What You Get

Before any credits are consumed, the upload tool validates your file — checking format, duration, and size. You know exactly what you're getting into before processing begins.

Technical Highlights

The stack is modern but practical:

Next.js 16 with App Router and Turbopack
AssemblyAI for transcription (Universal-2 model)
Cloudflare R2 for private file storage
Neon (PostgreSQL) for metadata and job tracking
Vercel for deployment
Better Auth for authentication
Creem for payments

One interesting detail: transcription jobs are processed asynchronously via webhooks. When AssemblyAI completes a job, it sends a webhook back to update the status. This keeps the UI responsive even for long files.

The Little Things That Matter

During development, I kept a running list of friction points from other transcription tools I'd used:

Timestamps that don't match the audio — Fixed by validating word-level timestamps from AssemblyAI
Speaker labels missing in exports — Speaker changes are included when the AI detects them
72-hour retention — Enough time to review and export, short enough to feel privacy-preserving
Mobile-friendly — The full workflow works on a phone browser

What I Learned

Building a transcription tool teaches you things about audio quality you never thought about. Clear speech in a quiet room produces dramatically better results than a conference call with echo. Background noise, multiple speakers talking over each other, heavy accents, and technical jargon all challenge the AI. This is why MP4Transcript shows you the transcript for review before you export — the AI is good, but it's not perfect, and being honest about that is part of building trust.

What's Next

The roadmap includes more export formats (DOCX support is planned) and integrations for creators who want to add captions to their videos automatically.

If you have feedback or feature requests, the contact page is live — I'd love to hear what's missing.

Try it free: New accounts get 15 free transcription minutes. No credit card required.

mp4transcript.com | GitHub

🚀 Why I Built a Smarter AI Birthday Video Maker

dragong — Thu, 18 Jun 2026 16:49:44 +0000

Smarter Birthday Video, Create Simpler Joy. — https://birthdayvideo.ai/

1. Why I built this?

Every time a close friend or family member has a birthday, we all want to send a heartwarming video blessing. But when trying existing solutions, I ran into the same frustrating walls: you either waste hours wrestling with clunky software only to get an unsatisfying result, get stuck with generic, cheap-looking templates, or find that mainstream AI tools fail to hit the right vibe due to a lack of birthday-specific prompt optimization.

A birthday video shouldn't just bring joy to the recipient; preparing it should be a simple, joyful experience too. I just wanted a way to effortlessly create a high-quality birthday video within minutes. Since I couldn't find a solution, I decided to build one myself.

2. Tech Stack & The "Dual-Engine" Design

To perfectly balance "rendering efficiency" and "AI imagination," I ended up designing two completely different video generation paths:

⚡ Engine A: Flash Template Mode (Powered by Remotion)

Built on an innovative, pure front-end Remotion tech stack that operates entirely independent of LLMs. Its core edge is speed — supporting real-time live previews and wrapping up the final render in under 1 minute. It delivers highly consistent, polished videos, tailored for anyone prioritizing efficiency.

🎨 Engine B: Infinite Creative Mode (Powered by Seedance AI)

Integrated with industry-leading video large models. The secret sauce here is that I have deeply optimized and embedded a set of highly efficient, implicit prompts tailored specifically for birthday scenarios. Without needing complex inputs, a minimal adjustment lets the AI unfold a highly personalized visual masterpiece within 3 minutes.

3. Product Details & UX Thinking

Single-Page Application: I absolutely despise nested forms and multi-step redirects. So I obsessed over the workflow to ensure that selection, previewing, generating, and downloading all happen seamlessly within a single page.
Native Vertical Aspect Ratio: Since 99% of birthday videos are shared and viewed on mobile-first platforms like WhatsApp, TikTok, and Instagram, the app defaults to a mobile-friendly vertical layout.
Smart RSVP Integration: Many users create these videos for actual birthday parties. To bridge the gap, I baked an RSVP collection feature right into the app, turning a simple blessing video into a seamless social hub.

4. Feedback & Hard-Hitting Reviews Wanted

My friends and I have already become daily active users of this little utility. As a newly launched solo indie project, I am incredibly eager to get some raw, honest, and hard-hitting feedback from the community:

From a UI/UX standpoint, is the interface intuitive enough for non-technical users?
How is the video generation and rendering speed experiencing on your end under different network conditions?
What other fun, creative ideas do you have for a birthday video? I’d love to hear them and see how we can bring them to life!

Give it a spin here: 👉 https://birthdayvideo.ai/

DEV Community: dragong