How to Balance Source Volume and Your Own Voice While Shadowing

Stop guessing shadowing volume percentages. Use a MODEL LOST / SELF LOST calibration loop so you can hear the source and still monitor your own speech.

Scarlett Vale

FunFluen editor · Film and listening

Sep 4, 2026 7 min

The short answer

The useful shadowing balance is not a fixed percentage: set the source so it stays trackable while you speak, but not so dominant that your own output disappears from awareness.

If your shadowing setup keeps turning into an arms race—audio louder, then you louder, then audio louder again—the solution is not a magic number. Balance by what disappears.

Start here: which voice disappears?

Use the result of one short shadowing pass to decide whether volume needs changing.
What happens	What it suggests	Next move
MODEL LOST	Your own speech masks the source enough that you stop tracking useful model cues.	Make the model easier to perceive without deliberately speaking louder. Change one setup variable, then retest.
SELF LOST	The source dominates so much that you stop noticing your own endings, rushing, or other output cues.	Reduce source dominance or change the listening setup, then retest at a comfortable speaking effort.
BOTH PRESENT	You can still track the model and notice something about your own speech.	Freeze the volume. If the pass still fails, volume is probably not the main bottleneck.

You do not need both voices to feel equally loud. You need both to remain available enough to do their jobs.

Why a louder source can quietly change your own voice

Your voice is not produced in a vacuum. Speakers monitor themselves through several feedback channels, including auditory and bone-conducted feedback as well as proprioceptive and tactile information. A review of speech self-monitoring discusses these multiple channels rather than treating your own voice as one simple external sound meter. See Lind and Hartsuiker’s review of speech self-monitoring.

That matters because changing what you hear can change how you speak. A scoping review of adult auditory-feedback studies found a broad pattern: when auditory self-feedback was masked, speakers tended to increase vocal intensity; when self-feedback was amplified, speakers tended to reduce it. The experiments were highly varied, so this is not an evidence-based instruction to use a specific playback level for shadowing.

It does explain why “just turn the source up” can become self-defeating. You may respond by speaking louder, which makes you raise the source again. Tiny arms race, now with extra vowels.

The Two-Channel Calibration Loop

This is a practical FunFluen framework, not a scientific volume formula. Use it on one line before a longer shadowing pass.

I didn’t expect the meeting to run this late.

Play the line without speaking. Set the source at a comfortable level where you can clearly track the stressed words and hear the end of the sentence.
Play it again and shadow at your normal comfortable speaking effort. Do not try to match the actor’s loudness.
During the pass, notice one model cue: perhaps the strong beat on expect, the rhythm of run this late, or the final word.
Notice one self cue: perhaps whether you swallowed the ending, rushed a phrase, or suddenly got louder.
Classify the pass as MODEL LOST, SELF LOST, or BOTH PRESENT.
Change one variable only, then replay the exact same line.

MODEL LOST: I hear myself, but the source disappears when I start speaking

Do not begin by shouting less subtly. Keep your speaking effort comfortable. Make the source a little easier to perceive or change the listening setup so the model is not buried, then replay the same line. If a small adjustment does not fix it, the clip or setup may be the issue rather than your volume control.

SELF LOST: I follow the source, but I notice almost nothing about my own output

Reduce source dominance and try again. Your goal is not to hear a studio-quality recording of yourself in real time; it is simply to retain enough self-awareness to notice something useful—an ending, a rush, a missing beat, a sudden increase in effort.

BOTH PRESENT: I hear the source and can notice my own speech, but I still fail

Stop moving the volume slider. You have already answered the volume question. The remaining problem may involve the material, timing, language knowledge, or another practice variable. Endless recalibration will not turn a different bottleneck into a volume problem.

The smallest useful test: one cue from each voice

After a pass, answer two questions:

Model: What did I hear the speaker do?
Self: What did I notice myself do?

A useful answer might be: The speaker stressed expect; I rushed the meeting. That is enough. You do not need simultaneous forensic analysis of every phoneme while speaking.

If you can only answer the model question, SELF is being crowded out. If you can only answer the self question, MODEL may be disappearing. If you can answer both but the line is still messy, volume has probably done its job.

Headphones can change the equation

Different listening setups change how much of your own voice reaches your attention. Highly isolating headphones can feel very different from a phone speaker or a more open setup. Some current headsets also offer a feature called sidetone or self voice, which routes some of your own microphone signal back to the headphones.

For example, Sennheiser’s current support documentation defines Side Tone as letting users hear their own voice through headphones and notes that availability and controls depend on the specific model and firmware.

That is an optional device feature, not a requirement for shadowing. Do not assume your headset has it, and do not assume someone else’s headset setting will transfer to yours. The calibration loop works from the result you can actually hear.

Copy the speech pattern, not the character’s loudness

If a movie character whispers, screams across a battlefield, or delivers a dramatic courtroom monologue, their loudness is part of the performance—not a target you must physically reproduce.

Separate useful imitation from volume competition.
Source feature	Useful to copy?	Your job
Stress and rhythm	Usually yes	Track the strong beats at your comfortable voice level.
Phrase boundaries and pauses	Yes	Follow the grouping without copying theatrical loudness.
Pitch direction	Useful when relevant	Imitate the contour without forcing the source’s intensity.
A shout or extreme whisper	Not as a loudness target	Keep the communicative pattern; use your own comfortable voice.

Do not out-yell an action hero in the name of pronunciation. The hero has a sound engineer. You have neighbours.

Re-test the same line, not a new one

Calibration gets muddy if every attempt uses different material. Once you change one variable, replay the same sentence and ask whether MODEL and SELF are both back.

On supported video pages with subtitles, FunFluen can reduce that playback friction with sentence navigation, repeat controls, and fine-grained playback speed when a slightly slower diagnostic pass is useful.

Review FunFluen for repeatable line-by-line shadowing practice

The first action is reviewing the extension listing before installing. FunFluen does not claim to mix your microphone voice against the source or control your headset’s sidetone; device audio behaviour varies.

Useful English for changing the volume

Volume control has its own little cluster of English phrases. One common learner pattern is understandable but unnatural:

Natural ways to talk about audio volume.
Learner wording	Classification	Listener interpretation	Natural alternative	Context note
Turn the volume more high.	Unusual/non-idiomatic in standard conversational English for this intended meaning	Increase the audio volume	Turn the volume up.	High volume is natural as a noun phrase: The music was playing at a high volume.

Turn it down is direct and natural when you control the device yourself. When asking another person, Could you turn it down a bit? is softer. For your next practice setup, say one instruction aloud before touching the control: I’m going to turn the source down a little and try the same line again.

Then make the original practice sentence yours: I didn’t expect the journey to take this long. That final change moves you from pure imitation toward usable speech.

The stop rule: once both channels are present, leave the volume alone

If those are true, stop tuning the mix. Continue the language exercise. If you want to place this drill inside a wider practice system, see FunFluen’s media-based language learning methods.

Neither voice needs to win

The source is there to lead. Your own voice is there to produce—and to give you enough feedback to notice what happened. Shadowing gets much harder when either one becomes invisible.

So forget the mythical perfect percentage. Run one line. If MODEL disappears, restore the model. If SELF disappears, restore self-awareness. If both are present, freeze the volume and get back to learning English.

Balance by disappearance, not percentage.