Best way to upload music to streaming sites likes YT, Soundcloud etc.?

[Pxz]

Is lowering the overall RMS level of the track a must before uploading? Do i need to upload in wav or mp3, i keep hearing different takes on this subject.

 

[Dans]

Check out this video from The Pro Audio Files which explain the whole thing in great detail. I basically follow suit by putting my final limiter output at -0.5 dB and use 16 bit dithering after that. I've heard about a technique where you narrow the stereo width for only the very high end so Soundcloud won't kill your cymbals but I just find it an uncomfortable thing to do lol.

 

[kapu]

Well, always upload in lossless format (e.g. 24-bit LPCM .wav). You can test how your sound survives different encoders. I believe SoundCloud encodes to 128 kbps MP3, YouTube encodes to 192 kbps AAC, Spotify uses around 100 kbps Ogg Vorbis for lowest quality and around 300 kbps for premium. iTunes uses 256 kbps AAC (at least for some playback profile).

If you're wondering about the 'optimal' loudness level, or RMS, I don't think at least SoundCloud loudness normalizes the material. Last time I measured the bitstreams Spotify normalized to +3 ReplayGain and YouTube to +4.5 ReplayGain for music. iTunes uses some arbitrary Apple invented algorithm, which is very close to around -16 LUFS Integrated. I'm not sure about Spotify Free, but Spotify Premium users get to disable the normalization. None these services normalize the material destructively, but provide metadata for the player software to forcefully limit the maximum playback level to match the normalization specification in use.

Main 'problem' with SoundCloud (and low quality Spotify) is of course the low bitrate. At lower bitrates audio is encoded using mid-side (M/S, Joint Stereo) method. Basically this means instead of equally dividing the bandwith ('bitrate') between left and right channels, the encoder uses most of the available bandwith for the mid (mono) part of the signal, and relatively smaller amount for side (the difference) channel. This narrows the stereo image a bit. If the mix has lots of difference between the channels at high frequencies ('stereo highs'), they will suffer and distort more noticeably.

Lossy encoding, such as MP3, requires some headroom. The louder (low RMS to peak ratio) the mix, more headroom is needed. This also depends heavily on the frequency spectrum of source material. Instead of adjusting the final master limiter threshold (or gain) it might be more convienent just to take the ceiling (output level) down a dB or two for the encoded print/export/render/bounce whatever it's called. Also, using oversampling (or intersampling) limiter will give more consistent peak levels, meaning the final true peak level after decoding the encoded file will be closer to the originally set limiter output ceiling. ^_^

 

[nezvers]

Soundcloud, YouTube, Spotify and iTunes doesn't use set normalization volume but use some kind perceived volume algorithm (each does their own - average dynamic range normalization, peak normalization, etc).

 

[Redwind]

Any tips for uploading to Bandcamp by chance?

 

[kapu]

Any tips for uploading to Bandcamp by chance?

Hi! Measured few songs from Bandcamp, and didn't find any common denominators between most popular loudness algorithms. Instead, all the individual songs were peak normalized to -3 dBFS for 'low quality' MP3 streams. I'd suspect you get the 'full level' when purchasing and downloading the higher quality files. So, I'd recommend to upload your lossless 24-bit or 16-bit masters 'untouched' to Bandcamp.

A relatively 'fool proof' advise is to get the loudest part of the single/album to -9/-6 dBFS RMS ('old RMS' [e.g. Logic Pro RMS meter]/AES-17 RMS [e.g. Reaper RMS meter]), so in most cases the overall master level will be in line with most of the contemporary pop/rock releases. Also, in this case practically all stream services normalize the record downward by at least 3 dBs thus providing the extra headroom needed for oversampled peaks which occur during the decoding process. And in most cases the playback equipment/software also provides internal headroom. Unless using ultra hifi equipment capable of streaming the source data untouched to the converters without any extra dsp, but in such cases using MP3s or other encoded audio as source material would be kind of... unorthodox.

In a technical sense there would be benefits of making use of the higher peak-to-rms ratio and optimizing masters for e.g. YouTube or Spotify, giving somewhat better transient response and more perceived depth. The main problem with this is that the 'collective ear' is so used to the lower microdynamic content in contemporary loud masters, that 'better' microdynamics is considered 'sharp and edgy' and perceived as 'unfinished', even by major label ARs. Even though there were some voices declaring that the so called 'loudness war' is now over, I haven't seen any indication yet. Actually, it's vice versa. Some of the latest international hit rap releases are now momentarily pushing 0 dBFS RMS and the master sounds clearly distorted, yet it doesn't pose any problem for the hundreds of millions of fans listening them every day. So, in the end it's probably just safer to conform the masters (in terms of frequency spectrum and loudness) to the 'mass' instead of trying to get that 'extra' benefit by making use of the every last bit available at the release platform. ^_^

 

[Redwind]

Hi! Measured few songs from Bandcamp, and didn't find any common denominators between most popular loudness algorithms. Instead, all the individual songs were peak normalized to -3 dBFS for 'low quality' MP3 streams. I'd suspect you get the 'full level' when purchasing and downloading the higher quality files. So, I'd recommend to upload your lossless 24-bit or 16-bit masters 'untouched' to Bandcamp.

A relatively 'fool proof' advise is to get the loudest part of the single/album to -9/-6 dBFS RMS ('old RMS' [e.g. Logic Pro RMS meter]/AES-17 RMS [e.g. Reaper RMS meter]), so in most cases the overall master level will be in line with most of the contemporary pop/rock releases. Also, in this case practically all stream services normalize the record downward by at least 3 dBs thus providing the extra headroom needed for oversampled peaks which occur during the decoding process. And in most cases the playback equipment/software also provides internal headroom. Unless using ultra hifi equipment capable of streaming the source data untouched to the converters without any extra dsp, but in such cases using MP3s or other encoded audio as source material would be kind of... unorthodox.

In a technical sense there would be benefits of making use of the higher peak-to-rms ratio and optimizing masters for e.g. YouTube or Spotify, giving somewhat better transient response and more perceived depth. The main problem with this is that the 'collective ear' is so used to the lower microdynamic content in contemporary loud masters, that 'better' microdynamics is considered 'sharp and edgy' and perceived as 'unfinished', even by major label ARs. Even though there were some voices declaring that the so called 'loudness war' is now over, I haven't seen any indication yet. Actually, it's vice versa. Some of the latest international hit rap releases are now momentarily pushing 0 dBFS RMS and the master sounds clearly distorted, yet it doesn't pose any problem for the hundreds of millions of fans listening them every day. So, in the end it's probably just safer to conform the masters (in terms of frequency spectrum and loudness) to the 'mass' instead of trying to get that 'extra' benefit by making use of the every last bit available at the release platform. ^_^

Thanks so much for the info! I have a habit of ABing my songs with others that I find on Soundcloud or Youtube, so it's really helpful to also know the differences between the platforms.

 

[Pxz]

Thanks guys , all more clear now.