Audio and the mix
How Dafado turns a raw recording into the short, produced clip you hear on a review page.
What you hear is a short mix
When you press play on a review page, you are not hearing the raw recording straight off the microphone. You are hearing a short mix that Dafado produces from it. The mix is made to feel present and produced without changing what the person said.
A published clip has three parts:
- A brief logo sound plays first, so you know you are hearing a Dafado voice.
- The speech, evened out in volume so one review is not far louder or quieter than the next.
- An optional music bed, a low, soft layer of music that plays underneath the speech.
The music sits well under the voice. It is there to warm the clip, never to compete with the person talking.
The raw recording is never touched
This is the rule the whole pipeline is built on. The original recording is kept exactly as it was made. The mix is a separate, produced version. Because the raw file is never altered:
- Nothing about the music or the leveling can change the words.
- The mix can be remade at any time with nothing lost.
The transcript reads the raw recording
When Dafado writes out the transcript, it listens to the raw recording, not the mix. Music and leveling never reach the part of the system that turns speech into text. So the transcript reflects exactly what was said, and the mix is purely about how the clip sounds when you play it.
The same everywhere
The mix is made once and stored. Whenever and wherever you play a review, you hear the same clip. It does not get re-rendered per visitor or sound different on one device than another. One recording, one produced mix, played back the same for everyone.