Maximizing Audio Quality For Your Music On Audius
We went deep into the Audius transcoding flow to understand how to maximize the quality of streaming playback. Here's how to take advantage on the highest quality, 100% free music streaming platform.
Audius offers the best quality free music streaming on the market today, but if you're uploading music to the platform, there's more you can do to ensure your fans get the best experience possible.
In short, all tracks are streamed back on Audius as MP3s with a constant bitrate of 320kbps, at a sample rate of 48000 (48kHz).
However, if you're not careful about what files you're uploading, you can still degrade the quality of your work. There's a couple simple things you can do to ensure your music keeps its glorious fidelity.
Potential Degradation Factor #1 - Audio Sample Rate
If you are exporting your tracks with a sample rate of 44.1kHz, the Audius transcoding pipeline will automatically resample your tracks up to 48kHz. However, the audio file you're uploading doesn't have this depth of information, so the transcoder will have to fill in the new samples so that DAC (digital audio converters) can reconstruct the audio at the new sample rate (implemented via bandlimited interpolation). Unfortunately, implementation of this process is not standardized across software, and resampling results vary from vendor to vendor. Measurable change in your signal is unavoidable when this is needed, but generally, any artifacts due to spectral imaging are often imperceptible.
What can happen during upsampling is the introduction of intersample-overs, or brief moments of clipping due to the resampling of your audio. When the audio is resampled at a higher rate, new samples are used to reconstruct your signal that are different than the ones you exported from your DAW. For loud, heavily compressed masters, DACs may reconstruct the signal in a way that cause the signal to clip, that you wouldn't have been able to catch prior to bouncing your audio at a lower sample rate. Since the devices and hardware listeners use to enjoy your music vary greatly, you may not even know its happening.
Similarly, resampling in the opposite direction has the chance to introduce artifacts when you supply a track with a sample rate above 48kHz. It's important to note that resampling artifacts that pollute the frequency spectrum most of the time are imperceptible, but all audio is different. It is possible for artifacts to become perceptible if your audio is unlucky enough to meet the set of conditions that would cause audible degradation such as aliasing, smearing of your stereo image, or loss of clarity in your bass.
Solution - Upload tracks that don’t require resampling:
If you’re worried about any of these issues affecting your music, the simplest way to be safe is to export your master with a sample rate of 48kHz. If you use this file, Audius’ transcoding flow won’t have to do any resampling when preparing your file to stream to listeners.
Potential Degradation Factor #2 - Uploading your master in a lossy compression format (MP3, M4A, lossy WEBM, etc)
When you upload your audio file to Audius, it will be re-encoded to a new MP3 no matter what, even if you upload an MP3. However, MP3 is a lossy format, so re-encoding an MP3 to an MP3, even with the exact same sample rate and bitrate will result in a loss of quality. This is similar to how each time you export a JPEG image to a new JPEG, the quality will degrade with each successive iteration.
If you've uploaded a 320kbps MP3 to Audius in the past, fear not, the loss in quality is minimal, but it is still there.
Solution
Upload your original audio track in a lossless format, (e.g. WAV, FLAC, AIFF). Avoiding the conversion to MP3 on Audius unavoidable, so the best way to minimize quality loss is to give the transcoder your lossless file.
Should I dither my lossless files I supply to Audius?
No. Dithering is not necessary when converting from a lossless to a lossy MP3 (or any other lossy format). The noise that’s introduce via dithering will not remedy any potential artifacts introduced during the lossy conversion process. Dithering is only relevant to minimize quantization artifacts when lowering bit depth on lossless tracks (for example, converting a 24bit wav to a 16bit wav).
Other considerations - The Limiter Ceiling on your master chain
When converting from a lossless to lossy format, sometimes the resulting file will peak above where you initially set the ceiling of your limiter on the master, resulting in clipping artifacts. To remedy this situation, a common practice is to set the ceiling on the final limiter in your master chain between -.3dB and -.8dB (though sometimes the track may require that you go lower).
Providing this extra bit of headroom on the lossless master is often enough to give the resulting lossy transcoded track the leeway it needs so that clipping won’t occur.
Audio Summary
For best quality, aim to upload audio files with the following attributes:
Lossless Format - WAV, FLAC, AIFF, etc
A Sample Rate of 48000 (48kHz)
No Dithering
Final Limiter Ceiling on your master chain between -0.3 dB & -0.8 dB
Bonus - Maximizing the Visual Quality of Your Cover Art Images
For your track, your cover art image will be converted to various size JPEGs, to display to users on Audius’ various desktop and mobile clients.
The largest of these will have dimensions of 1000x1000 pixels, so you should endeavor to upload an image no smaller than this. Otherwise, Audius will have to increase the size of your image, and upscaling images is a guaranteed way to create visible artifacts.
Similar to the audio transcoding flow, Audius will create a new JPEG image with a 90/100 quality setting using the source image you upload, even if you upload a 1000x1000 JPG. JPEG is a lossy format, so even if you're uploading a JPEG 100 (lowest possible compression setting), compression against your source has already occurred. If you’re uploading that image to Audius as your source file, you're introducing another lossy compression step before that image is ultimately displayed.
To minimize the number of destructive compression stages, you should export and upload an image in a lossless format, such as PNG. The displayed image on your artist profile will then only have been subjected to a single compression stage, resulting in your cover artwork being displayed with the highest possible quality.
Cover Art Image Upload Summary
For the best possible image quality for your Audius cover artwork, use:
An image with dimensions of at least 1000x1000 pixels
Exported in a lossless format
The Bigger Picture
Other free-to-stream or ad-supported services such as soundcloud are unable to match the streaming quality of Audius due to the larger file size of higher quality audio files. Audius is unique in its ability to stream back high quality audio for free and without ads as its supported by a decentralized network of Content Nodes, that facilitate the content delivery aspect of running a streaming service. Content Nodes are a key of the Audius protocol (together with Discovery Nodes) that comprise the Audius Protocol. By providing incentives for Node Operators, Audius is able to ensure high quality streaming as a default. See the following links for more in-depth information on the functioning of the protocol:
Audius Protocol Overview
https://docs.audius.org/token/running-a-node/introductionAudius Protocol Nodes Status Resources
https://audioprime.io/audius-discovery-nodes-status
https://dashboard.audius.org/#/services
If you’re an Artist, be sure to check out audioprime.io for data & analytics resources for your tracks and artist profile.
If you have feedback or thoughts, connect with us on twitter (twitter.com/audioprime). If you want to stay up to date with detailed information about new Audio Prime features, as well as discussions on Audius, we encourage you to subscribe to this newsletter.