AI Video Voiceover and Sound: 3 Fixes, Before and After
Three voice and sound fixes from a 90-second in-house sample, with before-and-after audio clips for the voiceover, one mispronounced word and the song.
Even when the visuals are well made, an odd voice or sound pulls the viewer's attention away from the message.
This post covers three voice and sound fixes we made before finishing "Cosmetics PR," a 90-second in-house sample, with before-and-after clips you can compare.
Turning a theatrical voiceover into natural narration
In the first cut, the voiceover lines were delivered with heavy emotion. The brand copy was fine, but every other line sounded theatrical.
When we analyzed the voice, we found that regenerating it had produced a rushed version with the pauses squeezed out. The six lines had shrunk from 15.5 seconds in total to 9.9 seconds. We went back to the previous version and fixed the lines one at a time with a flatter intonation.
First cut. Lines read with heavy emotion
"We made something good. That alone didn't reach anyone."
A lower, calmer voiceover
"We made something good. That alone didn't sell."
Checking the pronunciation of every word
In one version, the Japanese word "shohin" (product) in the line "convey the product's features" sounded like "shopin."
We left the on-screen captions in kanji and regenerated the audio from a voiceover script that spelled out the reading in hiragana, the Japanese phonetic script.
The version where "shohin" in "convey the product's features" sounds like "shopin"
The version with the reading specified as "shohin"
Keeping the song going under the voiceover
In the first cut of song option C, the vocals lasted only 16.5 seconds, and 0:41 to 0:50 was accompaniment alone.
For the final cut, we remade the song with lyrics running through the whole video and lowered the vocals by about 5 dB only while the voiceover plays. We also layered in cafe room tone, rain, footsteps and coffee sounds, which took the ambient sounds and sound effects from 17 to 27.
0:41 to 0:50. Accompaniment only, no vocals
"Why not just talk to someone first?"
Same section. The song keeps playing quietly, with cafe sounds
"Why not ask someone for advice first?"
Before (first cut of option C)16.5 sec of vocals, 17 ambient sounds and sound effects
Final cutVocals from 0:00 to 1:28, 27 ambient sounds and sound effects
Checking sound together with the picture
Voice and sound adjustments are less visible than fixes to the picture, yet they strongly shape how a video feels once it ends.
A comparison of the first and final cuts of the visuals, along with the instructions we gave during production, is on the video case study page.
Talk to us about your video
You don’t need a finished plan to get in touch.
ContactSee pricing and the production process
