Audio loudness

Videos from different cameras, editors and people have different loudness. Without a common level, a viewer turns the volume up for one video and down for the next. Kitset can give each new video the same level when it encodes it. Each site has a default level, and each upload can choose a different level.

Loudness in short

Three measures describe the level of audio.

• Loudness, in LUFS (loudness units relative to full scale). It measures how loud the whole video sounds to a person, not only its loudest moment. A higher number is louder. −14 LUFS is 10 units louder than −24 LUFS. One unit is the same as one decibel.
• True peak, in dBTP (decibels relative to true peak). It is the highest point of the waveform. Audio above 0 dBTP distorts. Kitset keeps every peak at −2 dBTP or lower, because the encoder can make a peak a little higher.
• Dynamics: the difference between the quiet parts and the loud parts. Compression makes this difference smaller. A small amount of compression is common in broadcast, but too much makes audio sound flat.

The three levels

Web

Web suits most sites. −14 LUFS is near the level of YouTube and of much other video on the web, so your videos sound as loud as the videos that people watch elsewhere.

Kitset measures the loudness and the true peak of the audio. Then it calculates two amounts of gain: the gain that brings the loudness to −14 LUFS, and the gain that brings the highest peak to −2 dBTP. It uses the lower of the two. So Kitset never compresses the audio or cuts its peaks to reach the target.

The result is that loud audio always comes down to −14 LUFS. Quiet audio with high peaks can stay below −14 LUFS, because more gain would make its peaks distort.

Broadcast

Broadcast brings the audio to −24 LUFS, with the peaks at −2 dBTP or lower. Kitset uses the two-pass loudnorm filter of FFmpeg, which follows the EBU R128 method.

When one gain cannot reach −24 LUFS within the peak limit, the filter changes to a dynamic mode and compresses the audio. This happens when the peaks are more than about 22 dB above the loudness, or when the range of the audio is very wide.

Broadcast is quieter than most web video. On a phone or a laptop, a viewer can need more volume. Choose it when your videos must match a broadcast standard, or when your site already has many videos at this level.

Keep Original Level

The audio keeps the level that it has in your file. Choose it when the audio is already mastered for the web, for example a music video, and you want no change.

Stereo that Kitset makes from surround audio is the exception. A downmix adds the channels together, and that can make the stereo louder than the source. So when the peaks of the downmix would go above −2 dBTP, Kitset makes it quieter, but it never makes it louder.

Examples

Choose the default of a site

Open Site Settings ▸ Audio, choose the Default audio loudness, then click Save in the bar at the bottom.

• Single uploads, bulk uploads and Vimeo syncs use the default.
• A new site starts with Web. A site from an earlier version of Kitset starts with Broadcast, because Kitset used that level for every video before the setting existed. So its new videos sound the same as its earlier videos.
• A change applies to the next encodes only. Published videos keep their level until you replace their audio.

All the Macs of a site use the same default, because the setting is part of the site.

Choose a level for one upload

The confirm screen of each upload has an Audio section with an Audio loudness menu. The menu starts with the default of the site, and it marks that level “– Default”. When you choose a different level, Kitset warns that the video can sound louder or quieter than the other videos of the site.

A video with no audio shows no menu. A bulk upload and a Vimeo sync give each video the default, with no choice.

When you replace the video of a published folder, the menu starts with the default of the site, not with the level of the old video.

See the level of a video

• The edit window of a video shows its Audio loudness, or “No audio”.
• The filter menu of the Library has a section for the level. Use it to find the videos that do not match a new default.
• File ▸ Export CSV… has the columns “Audio Loudness” and “Has Audio”.

Kitset records the level of each video that it encodes. A video from before Kitset recorded the level shows as Broadcast, because every earlier version used that level.

Change the level of published videos

To change the level of videos that are already on your site, re-import their audio. Re-import Audio Only is a choice in Bulk upload and in Vimeo sync.

• Kitset encodes only the audio from the source file, with the default of the site. The video files, the link, the details, the poster and the subtitles stay.
• The source must be the file of the same video. Its length must be within one second of the published video, and it must have the same kind of audio, stereo or surround. A video that fails a check stays as it was.
• In Vimeo Sync, Show ▸ Different Audio Loudness lists the imported videos with a level that is not the default of the site.
• A person who watches a video during the change can get a short playback error. A browser can keep the old version of a video for up to 5 minutes.

Stereo and surround

Kitset always makes stereo audio in AAC at 256 kbit/s. When the source has more than two channels, Kitset also makes 5.1 surround audio in Dolby Digital Plus (E‑AC‑3) at 640 kbit/s. Each player chooses the audio that it can play. Kitset sets the level of each one.

Kitset uses the first audio track of the source file, or the track that the file marks as the default.

Silence and very short clips

Some audio has no loudness that Kitset can measure: digital silence, audio that is quieter than −70 LUFS, and a clip that is shorter than 0.4 seconds. This audio keeps its level, and the video keeps its sound.

What a level cannot fix

• A level does not make speech clearer, and it does not change a mix.
• A large gain on quiet audio also makes its noise louder.
• Viewers still control the volume of their own devices.