Wikipedia:WikiProject Spoken Wikipedia/Recording guidelines

From Wikipedia, the free encyclopedia

Spoken Wikipedia
Recording process

PrepareRecord (Read) → UploadAddNotify

Step 1: Record the article[edit]


  • Use the Ogg Vorbis audio format. 44.1 kHz sampling rate, with one channel (mono) and Ogg quality 1 (about 70 kbit/s). If your recording software does not support Ogg, free Ogg Vorbis conversion applications are readily available, some of which are here.
  • Add a half second (500 milliseconds) of silence at the beginning of your sound file. This will eliminate buffering 'hiccups', which may cause the first syllable in your spoken article to be inaudible as the file begins to stream (play).
  • Properly format the filename. The filename must be prefixed by "En-" to show that the recording is in English, followed by the article title in canonicalized form. Finally the suffix "-article" plus the ogg extension. An example would be: En-Sample lemma-article.ogg
  • Treat links like text when reading. Any vocal indication of every link would disrupt the flow.
  • Speak more slowly than normal, by about 25%. You should have about 150–160 words per minute so people can comfortably hear an article. The most conversational speech is at 200 words per minute or more.
  • Tell people where it's from. Begin your recordings with:

    "Article name, from Wikipedia, the free encyclopedia, at E N dot wikipedia dot org."

  • Read out all headings as you come to them like this: "Section one: History", "Section two: Modern uses" and so on. To make it easier to remember the numbering, go to Preferences, and under Gadgets at the Testing and development section check the box Auto-number headings. Keep in mind that this preference tab doesn't exist at your GlobalPreferences list and that this section or the specific gadget may not exist on all localizations and/or sister projects.
  • Tell people how it's licensed. Wikipedia articles are licensed under the Creative Commons Attribution-ShareAlike 4.0 Unported License. Spoken versions are derivative works, so end your recordings with:

    "This sound file and all text in the article are licensed under the Creative Commons Attribution-ShareAlike 4.0 International License, available at creative commons dot org slash licenses slash by hyphen S A slash four point zero."

Production tips[edit]

  • Warm up. The vagaries of the day can leave your vocal cords clogged, especially if you have recently been eating or drinking milk-based foods such as cheese and chocolate. Some gentle humming before starting to record can cleanse and moisten the vocal cords leading to a better sound to the voice. Take a look at these vocal warm-ups for more ideas.
  • Stay relaxed. Whenever you realize that your voice is becoming tenser than normal, or you're having more and more trouble speaking fluently, take a break, and resume when you feel muscular tension going away. Keep some fresh (not cold) water in reach, and drink some whenever you feel your mouth drying out. Also, speak outward, with your chin up. Posture and facial expressions can affect voice recordings.
  • Record at a quiet place to reduce noise, in addition to digital noise removal (see next section). A long cord from microphone to computer may be useful in order to reduce the computer noise. A laptop is generally quieter than a desktop computer.
  • Try standing up. If you think your voice sounds a little thin or lacking in expression, standing while recording might help. Radio announcers and actors regularly stand while speaking to lend additional depth and confidence to their performance.
  • Speak across the microphone or to the side of it, rather than directly into it, to avoid pops and breath noise. A distance of several centimeters or inches away from the microphone is recommended. A pop shield may also be used to minimize hard consonants, and a de-essing processor may be used to remove murmured voices. It is possible to create your own pop shield using a wire hanger and a pair of tights/pantyhose.
  • Volume levels — make sure your audio levels are high but not clipping and, if you can, compress and normalize your audio.
    • For those unfamiliar with "audio" terminology, here's a brief description of these terms:
      • Compression is a dynamic leveling of audio, making loud parts quieter, and the quiet parts louder, so that a consistent sound level is achieved (on professional audio gear where 0 dB is maximum, –12 dB is a good place to level the average RMS for speech). Beware of excessive compression, as it will make noise stand out (even after noise reduction) and could exaggerate some sounds of speech.
      • Normalization is a calculated adjusting of audio so that the loudest peak is set to maximum potential volume, generally it is close to 0 dB (on professional audio gear where 0 dB is maximum, −10 dB is a good place to normalize the average RMS for speech).
      • RMS — see Audio power
  • If you mess up a take, don't stop the recording. Instead, just stop speaking and then speak out "three, two, one". Then start reading again. You can easily edit out the "flub" since the countdown gives you a cue on where to make the edit, and gives you some "silence" to edit in.

Step 2: Edit the audio[edit]

Remove noise[edit]

You should remove background noise from all your recordings. Here is a guide to doing this in Audacity:

  1. Make your recording.
  2. Select a chunk of the recording where you were not speaking. You should see a slight bumpiness on the line, representing the background noise.
  3. Select Effect, then Noise Reduction, then Step 1 and then Get Noise Profile.
  4. Select the entire recording (shortcut key: Ctrl + A).
  5. Go to Effect, then Noise Reduction and then Step 2.
  6. Drag the slider a little to the left, towards Less.
  7. Select Remove Noise.

The reason for moving the slider is that the default setting is rather powerful, and is likely to affect the sound quality by removing too much noise. When in doubt, remove too little rather than removing too much. A clue that you might have removed too much noise is hearing "bubbles" ("speaking in a glass"-like sounds) in the recording. If so, go to Edit and then Undo and try again with a lower setting.

Even at the lowest setting, this tool should remove all audible background buzz. If it is not able to deal with the background noise, then you probably need to record it again, this time without buses and motorbikes going by your window!

If all this seems like a lot of hassle, then send a message to a fellow contributor and they'll do it for you.

Adjust sound levels[edit]

You should be adjusting the recording volume so that the peaks you see on the screen do not touch the top and bottom when you speak. If they do, then "clipping" will occur (part of the sound information will be lost) and you'll just get a blaring noise. On the other hand, the volume should be set fairly high.

Once you have finished the recording, you should boost its volume. If you perform normalization as described above, the volume will increase. Normalization will amplify the signal as much as possible without causing "clipping". If normalization does nothing, then the software can't boost your voice without clipping. In this case, the chances are that you set the recording volume too high. It is also possible that there is a peak at some point in the recording because you made a loud noise (common culprits include shouting the first few words of a sentence, coughing, clicking the tongue, choking on food, scraping your chair legs on the ground). You should try to edit that out, or perhaps just have another try.

It is also possible to select individual parts of your recording and boost them separately. However, compression (part of the dynamics processing effect in Adobe Audition) does this automatically and dynamically on a far finer scale. Compression is preferable.

Those users who find the traditional process of compression, normalization and limiting as challenging could use Levelator instead, which works with uncompressed audio (file formats WAV and AIFF). While recording, after noise removal and cleanup, save or export the file in WAV format. Run Levelator. When the program is running, it only displays a large icon or splash-screen. Drag your WAV file onto the icon of Levelator. The file will automatically generate an output file in the same format. You can convert this output file to ogg vorbis, by importing it into Audacity and exporting it in ogg vorbis format. Levelator will adjust the audio levels within an audio segment (as opposed to traditional compression, normalization and limiting) by creating a new copy of the audio sample which has balanced levels and a standard overall volume.

Tips and tricks[edit]

Recommended tools[edit]

Use audio recording software that can record section by section, like:

Sample configurations[edit]

Many users use different sample configurations to record articles.

See also[edit]