Dev.to WebDev πŸ›  Dev πŸ‘ 0 πŸ“– 3 min read

Why I Built an Ad-Free, 4K Video Downloader in Next.js 14 & Python (And Killed Popups Forever)

Have you ever tried downloading a public video from Reddit or an educational lecture from YouTube recently? You type "video downloader" into Google, click on one of the top results, and within 3 seconds: Three sketchy

Have you ever tried downloading a public video from Reddit or an educational lecture from YouTube recently?

You type "video downloader" into Google, click on one of the top results, and within 3 seconds:

  • Three sketchy popups open in background tabs.
  • You see 4 blinking green "DOWNLOAD NOW" buttons (3 of which are adware installers).
  • You finally download the Reddit video, only to discover it has no sound because Reddit serves video and audio on separate streams.

As a full-stack engineer, I got sick of this hostile user experience. The web shouldn't feel like navigating a minefield just to grab an offline copy of a creative commons video, a transcript, or a TikTok tutorial without a giant watermark plastered over it.

So I spent the last few weeks building GetVideo β€” an ad-free, high-performance public media utility built with Next.js 14, FastAPI, Redis, and Caddy.

Here is a look behind the scenes at the architecture, the technical hurdles, and why modern web engineering makes this 10x faster.

The Tech Stack at a Glance

  • Frontend: Next.js 14 App Router (Server Components, Tailwind CSS, next-intl for English + Urdu RTL, PWA with native Web Share Target).
  • Backend API: FastAPI (Python 3.13, Pydantic v2 async validation, sliding-window token bucket rate-limiting).
  • Queue & Workers: Redis 7 + ARQ worker pool for asynchronous extraction and stream multiplexing.
  • Multiplexer: Custom FFmpeg stream-copy pipeline (-c copy).
  • Reverse Proxy: Caddy 2 with automatic Let's Encrypt TLS and HTTP/3 support.

Live application: https://getvideo.techhowl.site

Challenge 1: The Reddit "Muted Video" Problem (Zero-Transcode Stream Copy)

Most legacy downloaders fail on Reddit because Reddit uses DASH (Dynamic Adaptive Streaming over HTTP). The video stream (DASH_1080.mp4) and the audio stream (DASH_AUDIO_128.mp4) live on completely separate URLs.

Cheap converters try to run full FFmpeg re-encoding (libx264 + aac), which melts their server CPU and takes 30–60 seconds per video.

Instead, we use FFmpeg Stream Copy:

ffmpeg -i video_input.mp4 -i audio_input.m4a -c copy -map 0:v:0 -map 1:a:0 output.mp4

Because -c copy simply repacks the packets into a clean MP4 container without touching the underlying bytes:

  • CPU utilization is almost 0%.
  • Merging a 1080p 3-minute video takes less than 400 milliseconds.
  • Users get a pristine, audio-synced video instantly.

Challenge 2: Real-Time SSE Feedback (No Fake Spinner Animations)

Instead of a generic loading spinner, we implemented Server-Sent Events (SSE) between FastAPI and the Next.js client.

When a user submits a URL on GetVideo, the job is queued in Redis:

  1. queued (Queue position calculated)
  2. fetching (Extracting adaptive streams)
  3. merging (Instant FFmpeg stream-copy)
  4. ready (Signed single-use HMAC token generated)

The frontend renders a pleasant real-time progress bar. Once ready, the browser automatically triggers the file download via an ephemeral signed URL that self-destructs after 15 minutes.

Challenge 3: TikTok Without Watermarks & Bilingual Subtitles

Two features I really wanted for myself:

  1. TikTok No-Watermark: Content creators constantly need to repurpose their own TikTok videos to YouTube Shorts or Instagram Reels. We parse direct CDN streams to deliver 100% clean, uncompressed video.
  2. Subtitles & Transcripts Studio: Many people need transcripts for research or language learning. We built a tool into GetVideo that extracts .srt and .vtt tracks, exports clean plain-text transcripts, and can even merge bilingual subtitles (e.g., Spanish + English side-by-side).

You can test the subtitle extractor and tools directly on the live platform:

Privacy by Design (Ephemeral Storage)

We don't want your data. We don't store your videos.

  • We do not require accounts or logins.
  • Download tokens expire in 15 minutes.
  • A dedicated cleanup daemon runs every 15 minutes to permanently purge temp media from server storage.
  • Zero tracking scripts or sketchy third-party ad networks.

What’s Next?

I’m currently rolling out a lightweight browser extension and expanding platform extractors.

I'd love for the dev community to try it out, stress-test it, and share feedback!

Check it out here: https://getvideo.techhowl.site

What tools do you usually use for media extraction, and what features would you like to see next? Let me know in the comments!

πŸ“° Read the original article on Dev.to WebDev

Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes β€” full credit and traffic to the original publisher.