Published by PitchChanger · Updated · How these guides are made
How to use this workflow
Mark a slightly wider candidate
Set the end after the final consonant or room decay. The millisecond fields describe one contiguous interval and round to audio frames.
Preview the last phrase
Listen from before the final word through the endpoint. Move the end later if a consonant, breath, or intended tail disappears.
Apply a restrained fade if needed
Try a short linear, exponential, or S-curve fade-out, generate a fresh WAV, and check the saved ending.
Listen before and after the proposed endpoint
A waveform shows amplitude, not whether a listener needs the final “t”, “s”, breath, or reverb tail. Start with context before the final word and audition through the intended ending.
For a 12.340-second candidate endpoint, compare 12.240–12.440 seconds rather than judging the cursor alone. The numbers are an editing procedure, not a claim that every spoken ending needs 100 milliseconds of context.
Swipe or scroll sideways to read all columns.
| What you hear | Boundary decision | Fade decision |
|---|---|---|
| Final consonant is missing | Move the end later | Do not use a fade to hide a cut word |
| Word is complete; abrupt click remains | Keep the boundary | Try a short fade-out |
| Room tail is intentional | Keep more tail | Use little or no fade |
A fade cannot restore removed speech
Fades affect samples inside the selected range. They can soften a discontinuity but cannot bring back sound that lies after the end selection. Long fades can make the final syllable less clear.
Fade-in and fade-out are independent, and overlapping fades multiply. Generate again after changing selection or fades; an older download no longer represents the displayed settings.
Use the right editing tool for the task
Audio Trimmer keeps one continuous interval and exports WAV. It does not auto-detect silence, snap to zero crossings, make several ranges, or export MP3 directly.
To remove a mistake in the middle or join separated passages, use Audio Cutter. Convert the saved WAV afterward only when an MP3 delivery copy is actually required.
A boundary worksheet for a quiet final consonant
Suppose a word’s vowel ends at 2.340 seconds, its quiet final consonant continues to 2.460, and the next speaker begins at 2.800. An endpoint at 2.500 preserves the consonant plus 40 ms of context while excluding the next speaker. At 48,000 Hz it resolves to frame 120,000. These are constructed timing landmarks, not a recorded voice measurement.
A 200 ms fade ending at 2.500 starts at 2.300, before the consonant has finished. For a linear fade, the gain at 2.460 is only 0.2. Moving the end alone does not guarantee clarity: compare with no fade, then shorten the fade or add more context when needed. Listen to the whole word in the downloaded WAV, not just the visible waveform.
Swipe or scroll sideways to read all columns.
| Landmark | Frame at 48 kHz | Editing implication |
|---|---|---|
| 2.340 s | 112,320 | Vowel ends; consonant may remain |
| 2.460 s | 118,080 | Example consonant ends |
| 2.500 s | 120,000 | Candidate endpoint |
| 2.800 s | 134,400 | Next speaker begins |
Common questions
Can the tool detect the best word boundary?
No. It provides a manual one-range selection; listen to the candidate ending.
Will a fade fix a clipped final syllable?
No. Move the end later first. A fade only changes audio that remains selected.