A Million Ads delivers dynamic creative audio adverts and this guide is designed to help script writers and audio producers create for this format.
Our technique combines the script and corresponding audio with the data available about each user to create and traffic the most relevant and context aware ad for that user.
This is a general guide with some tips. Please get in touch if you have any more specific questions: firstname.lastname@example.org
The basic process
Creative can be personalised using many types of data
Dynamic creative ads are based on a script template
Write one complete version of the script
Be mindful of the data available
Highlight where the data influences the script
Avoid short fragments or mid-sentence chops, natural breathe points make good phrase edit points
Cover every option, always remember the default (i.e. what if no data is available?) Also check that both the longest and shortest combinations fit nicely in the time
Producer's Guide - The Recording Session
Record clips of the same element to approximately the same length. ±0.5s is normally fine
Start with the longest combination of alts for timing and pace before recording the alts
Avoid mid-sentence or consonant-to-vowel (or vice versa) edit points between the dynamic elements of the script
Make sure the voice over reads every alternate version with exactly the same intonation
Producer's Guide - Mixing the session
Make one complete version of the ad in ProTools (or Logic etc.) and get it sounding perfect for that version.
Drop the rest of the voice / sfx / music clips onto the relevant tracks so that the track effects are applied to them as well
Solo each track one-by-one and bounce out the result
Trim each clip to remove any silence on the top and tail
Have each clip individually and name the file as close to the line in the script as possible so our auto-matcher can automatically match it up.
Producer's Guide - Audio Specification
16bit 44.1 kHz Mp3 - at least 192 kbps (although WAV format is also accepted)
A Million Ads Studio handles the mix-down so level setting and any processing must be applied to the individual elements as would happen in a standard stem mix.
Levels - Best practice
Voice elements: - 12 dBFS peak or around -21 LUFS short term Music bed: -20 dBFS peak or around -32 LUFS short term (or whatever works best with the voice elements)
Bear in mind there is no processing on the final mix (apart from normalisation to -0.2 dBFS + what ever processing is applied by the publisher) so all creative must sound great without relying on master channel limiting / eq / compression.
Producer's Guide - Delivery of Elements
Voice elements need to be supplied ‘pre-mixed’ - ie. they are the correct level relative to any music beds or sound effects.
All voice elements should be fully finished with any processing / fx already applied.
If there are spot sound effects that link to certain voice lines, it is usually a good idea to mix these into the relevant voice elements (as long the sound effect does not extend to beyond where the voice finishes)
Music / Sound Bed Elements
Music elements such as beds should be supplied ‘pre-mixed’ so that their levels work with the levels of the other voice and sound effect elements (although it is possible to trim levels once audio has been loaded into Studio)
Sound Effects / Atmos Beds
As per music elements, sound effects should be ‘pre-mixed’ so that their levels work with the levels of the other voice and music elements (although it is possible to trim levels once audio has been loaded into Studio)