DEV Community

Cover image for Why 90% of YouTube to MP3 Tools Give You 128kbps When You Asked for 320
Hitesh Meghwal
Hitesh Meghwal

Posted on

Why 90% of YouTube to MP3 Tools Give You 128kbps When You Asked for 320

#python #webdev #opensource #tutorial

I Tested 8 YouTube to MP3 Tools. 7 Lied About the Bitrate.

You click "320kbps" on a YouTube to MP3 site. You wait. The download finishes. You play the file in VLC.

The bitrate? 128kbps.

You weren't exactly lied to — you were just handed a download from a tool that doesn't understand how YouTube serves audio in the first place.

I figured this out the hard way while building an audio extraction tool. Took me a week of debugging before I realized the problem isn't the output bitrate — it's how YouTube structures its audio streams at the source.

Here's what's actually going on.

YouTube Doesn't Store MP3 Files

This is the first thing most tutorials skip:

YouTube doesn't have MP3 files anywhere on its servers.

YouTube serves audio in two formats:

  • AAC (m4a) — used in MP4 streams, typically up to 128kbps (format 140), occasionally 256kbps (format 141, rare)
  • Opus (webm) — used in WebM streams, up to ~160kbps (format 251)

When you "download YouTube to MP3," what's actually happening is:

  1. The tool downloads one of those source streams
  2. FFmpeg transcodes it to MP3 at whatever bitrate you asked for

If the source stream is 128kbps Opus, transcoding to "320kbps MP3" doesn't recover quality that was never there. You get a real 320kbps MP3 file, but the audio inside can't exceed the fidelity of the 128kbps source.

It's like upscaling a 480p video to 4K — bigger file, same actual detail.

This is why so many "320kbps" downloads sound identical to 128kbps versions. Because the audio information inside is 128kbps, just wrapped in a bigger container.

The Format Codes That Actually Matter

When yt-dlp lists available formats for a YouTube video:

yt-dlp -F "https://youtube.com/watch?v=..."
Enter fullscreen mode Exit fullscreen mode

You'll see something like:

ID  EXT   RESOLUTION  ACODEC  ABR  NOTE
139 m4a   audio only  mp4a    49k  audio only, low
140 m4a   audio only  mp4a    129k audio only
251 webm  audio only  opus    160k audio only, high
Enter fullscreen mode Exit fullscreen mode

The ABR column is what actually matters — that's the average bitrate of the source.

Format 251 (Opus, 160kbps) is the highest audio quality YouTube offers for most videos. Some music content has format 258 (384kbps AAC) or 327 (256kbps AAC), but these are rare and mostly appear on YouTube Music URLs.

Most "free" tools just grab format 140 because it downloads fastest. They then re-encode to "320kbps MP3" to look impressive in the UI. Result: technically a 320kbps MP3 file, but with 128kbps source quality inside.

Out of 8 tools I tested, 7 did exactly this.

What Actually Works

If you want maximum quality YouTube audio, here's the right yt-dlp config:

ydl_opts = {
    'format': 'bestaudio[acodec=opus]/bestaudio',
    'postprocessors': [{
        'key': 'FFmpegExtractAudio',
        'preferredcodec': 'mp3',
        'preferredquality': '320',
    }],
}
Enter fullscreen mode Exit fullscreen mode

The key is bestaudio[acodec=opus]/bestaudio:

  • First preference: best Opus stream (usually 160kbps, sometimes higher)
  • Fallback: best audio of any codec

Without that filter, yt-dlp sometimes picks an m4a 128kbps stream because it scores "best" by some criteria.

For the MP3 conversion, preferredquality: '320' is the OUTPUT bitrate. A good tool should also detect and show users the actual SOURCE bitrate before download — so a video with only 128kbps Opus source isn't marketed as "320kbps."

The HLS Trick Most Tools Miss

YouTube uses HLS for live streams and some music videos. HLS is segmented — the audio is split into hundreds of 6-second chunks.

A naive downloader hits the manifest URL once, gets a tiny m3u8 playlist file, and gives up.

The fix is to download all chunks and concatenate them:

ydl_opts = {
    'format': 'bestaudio',
    'hls_use_mpegts': True,  # critical for HLS
    'postprocessors': [{
        'key': 'FFmpegExtractAudio',
        'preferredcodec': 'mp3',
        'preferredquality': '320',
    }],
}
Enter fullscreen mode Exit fullscreen mode

hls_use_mpegts tells yt-dlp to handle the segmented stream properly. Without it, you'll either get an empty file or a download that fails after 6 seconds.

What a Good Tool Should Actually Do

After all this debugging, here's what I think any honest YouTube audio tool needs:

  1. Detect the actual source bitrate first — analyze before showing options
  2. Show realistic quality choices — if the source is 128kbps, don't offer "320kbps" as an option
  3. Prefer Opus when available — for music tracks, this often unlocks higher bitrates
  4. Be honest about what you're getting

"Here's the highest quality YouTube serves, transcoded to MP3" is a better message than "320kbps PREMIUM HD AUDIO!" when the underlying source is 128.

Edge Cases That Will Bite You

A few more things I learned the hard way:

Age-restricted videos require login. yt-dlp has a --cookies option, but this is a maintenance nightmare in production. Better to show users a clear "this video requires login" message than fail silently.

Live streams have weird audio. YouTube live streams are HLS but with rolling buffers. You can't always download the "whole" stream — only what's currently in the buffer. Tell users this clearly before they expect a 3-hour podcast download to start instantly.

YouTube Music has different format codes than YouTube. A track on music.youtube.com may expose higher-quality streams than the same track on youtube.com. Worth checking both URLs.

Some videos have NO audio stream at all. Sounds impossible, but YouTube has plenty of silent meditation videos and visual-only content. The bestaudio format selector returns nothing — handle that case explicitly or your tool will crash.

What I'd Tell My 2-Months-Ago Self

I spent 2 months on this project. The audio quality problem alone took a week to fully solve. Three lessons from that week:

1. Don't trust UI labels in tools you don't control

I tested 8 different YouTube to MP3 tools while researching. 7 claimed "320kbps." 7 gave me 128. I almost gave up because I assumed they were doing something I wasn't. Turns out, they were just lying.

2. Read the format docs before assuming you understand the pipeline

I assumed "320kbps MP3" meant 320kbps quality end-to-end. It doesn't. It means a 320kbps MP3 container, which can hold any quality of audio inside, capped by the source. Reading the yt-dlp format documentation cleared this up — but I should have done it on day 1, not day 5.

3. Honesty beats marketing

Showing users "highest available: 160kbps Opus → transcoded to 192kbps MP3" sounds worse than "320kbps PREMIUM AUDIO!" but builds trust. People come back to tools that don't BS them.

If you've built audio extraction tools and run into similar issues — or you have a YouTube quirk you've battled — I'd love to hear about it in the comments.

I built AllClip while figuring all this out. Open to feedback if you want to poke at it.

Top comments (0)