Why Does Audacity Repair Reject the Damaged Sound in My Product Narration Even Though the Glitch Is Short?

Measure the selected samples, not the apparent duration

Repair estimates replacement audio from neighboring samples. It is designed for very short damage, such as a click or tiny dropout, rather than reconstructing missing dialogue. Audacity rejects selections longer than 128 samples.

At 48,000 samples per second, 128 samples represent approximately 2.67 milliseconds. At 44,100 samples per second, they represent approximately 2.90 milliseconds. These figures are arithmetic conversions of the documented limit, not evidence that every defect shorter than that duration can be repaired successfully.

Zoom into the waveform until individual samples are visible. Use the Selection Toolbar's sample-based display and Start and Length controls, as documented by Audacity, to identify the shortest interval containing the defect. A selection that includes an entire syllable will generally be far too long for this operation.

Check neighboring audio and clip state

Repair also needs usable audio outside the selected region on at least one side. Audacity uses those neighboring samples to estimate the damaged section. A small isolated fragment copied into an otherwise empty track may fail even when the fragment is fewer than 128 samples, because the surrounding information is missing.

The manual further states that Repair cannot be used on time-stretched clips. Check whether the narration has been retimed before diagnosing a disabled or rejected command. Preserve the original source and avoid assuming that flattening or changing the clip state will restore missing speech.

If the defect is a recording dropout that removes a complete consonant or word, interpolation from adjacent samples cannot reliably reconstruct the original language. Obtain another approved take or use an explicitly appropriate dialogue-editing workflow instead.

Hypothetical digital click at a product name

Hypothetical illustration—not an executed repair: a 48 kHz narration contains a digital click between the words 'model' and 'K4.' The editor initially selects 480 samples around the click, representing 10 milliseconds. Repair rejects that selection because 480 exceeds the documented 128-sample limit.

After inspecting the waveform, the editor identifies a hypothetical 60-sample damaged interval, or 1.25 milliseconds, with intact samples on both sides. This selection is within the numerical limit and has the neighboring material required for an interpolation attempt. It is not guaranteed to sound natural; the editor would still need to process a duplicate and listen.

A different fictional recording contains a missing 150-millisecond portion of 'K4.' Repair is not an appropriate reconstruction method for that loss. The correct decision is to find an authentic alternate recording or revise the approved narration plan.

Listen beyond the repaired point

Audacity recommends playing from a few seconds before the repaired region. This helps reveal clicks, discontinuities and changes that may not be obvious when auditioning only a tiny isolated selection.

If a first attempt sounds wrong, undo it and reconsider the exact sample boundaries. Do not widen the selection beyond 128 samples or treat the limit as a target that every repair must reach. Shorter, correctly localized defects are generally more suitable.

For product narration, the final test is the full spoken phrase, especially a model name, size or offer condition. Repair's availability establishes only that the sample-level operation is possible; it does not certify that the intended word remains accurate or understandable.

Record your own observations

The filled rows are illustrative examples. Use the blank row for your own files and observations.

Download the editable CSV worksheet

What the references establish

• Repair — Audacity Manual: Directly documents the 128-sample limit, neighboring-audio requirement, time-stretched-clip restriction, sample-level selection method and contextual listening review.