Ellie Lab

Lesson 04

TTS & music mix

Make the words easy to understand, then decide what music adds to the story.

Kitchen spill scene that needs clear dialogue over soft music
The kitchen sequence gives the sound a clear job: support the action while keeping every spoken line understandable.

TTS turns a script into speech. Ellie used Kokoro for the voices and FFmpeg for audio processing. You can study the audio decisions here without installing either tool.

Prepare speech one beat at a time

  1. Read one line from your script and check that its meaning is clear.
  2. Record or generate that line using the intended speaker and delivery.
  3. Listen for pronunciation, pace, clipped beginnings, and abrupt endings.
  4. Keep the best take and note its duration before moving on.

Separate takes make it easier to replace one awkward line. When joining them, listen to the transitions as well as the individual recordings. Two good takes may still need a pause between them.

If you are planning only

Read the practice script aloud and mark a short pause after each beat. Write “record again” beside any phrase that is unclear. That listening plan is the output for this lesson; creating audio files is optional.

Build a listening checklist

  • Words: can you understand the line without reading its caption?
  • Character: does the delivery fit the feeling of this beat?
  • Continuity: does the speaker sound consistent across takes?
  • Timing: is there enough space for a reaction or visual discovery?

Keep an untreated copy of the speech if you edit it. A later music change should not require recording the words again.

Add music with a purpose

Music can suggest warmth, tension, or discovery. It can also compete with a quiet voice. Begin with dialogue alone, then add a separate music layer at a low level. Bring it up only until it supports the scene.

Ellie uses Through The Woods by Shane Ivers. The music credits page provides the source, attribution, and CC BY 4.0 license link. If you choose another track for a finished video, record its own source and usage terms. Silence is also a valid creative choice.

A mix plan you can use in any editor

  1. Arrange the approved speech takes and pauses in story order.
  2. Place optional music beneath them on a separate track.
  3. Trim the music to the sequence; if it repeats, listen carefully at the join.
  4. Add a gentle fade at the beginning and end where it suits the scene.
  5. Reduce the music wherever it masks a word.
  6. Replay at an ordinary listening volume on headphones and a phone speaker.

Why there is no universal volume setting

Recordings and music tracks arrive at different loudness levels. Copying one numeric gain value will not produce the same balance for every pair of tracks. Judge whether the words remain clear throughout the scene.

Activity · Describe the sound

For each of your four beats, note the speaker, the intended mood, any pause, and whether music should be present. Explain one place where you would lower or omit the music. If you have an editor, compare a short preview with and without music.

Checkpoint

You can explain how the sound supports the story, identify a line that needs a clearer reading, and keep the music credit with your plan. A listener should not need captions just to overcome loud background music.