> ## Content Index
> Fetch the complete content index at: https://izotoperx.net/llms.txt
> Use this file to discover other available public pages before exploring further.

# How to Read the Spectrogram in iZotope RX 12
- URL: https://izotoperx.net/how-to-read-izotope-rx-spectrogram/
- Published: 2026-09-03T13:05:04.000Z
- Updated: 2026-09-05T01:04:12.000Z
- Description: Read RX from left to right for time, bottom to top for frequency, and dark to bright for intensity—then confirm every visual clue with your ears.
- Author: Brandon Hayes
- Tags: Getting Started, RX 12, Spectral Editing

Read an RX spectrogram in three directions: left to right is time, bottom to top is frequency, and dark to bright is intensity. A long horizontal line is usually tonal. A short vertical line is usually transient. A cloudy field of speckles can be broadband noise, but wanted sounds such as cymbals and sibilants can look noisy too. Those shapes are clues, not verdicts—play the selection before deciding what it is.

The spectrogram becomes useful when it answers two questions: where is the problem, and which frequencies does it occupy? A waveform still wins for overall level, timing and clipped peaks. RX lets you blend both displays, so you do not need to choose one view for every repair.

This walkthrough uses the standalone RX 12 Audio Editor, included in Standard and Advanced. RX 12 Elements supplies plug-ins without this editor; see the [RX 12 edition comparison](https://izotoperx.net/izotope-rx-elements-vs-standard-vs-advanced/) if your screen does not match. Open a working copy of your recording and turn **Instant Process off** before drawing selections: when it is enabled, a new selection immediately applies the chosen processing module.

## The three dimensions of an RX spectrogram

| Dimension | Where to read it                                | What it tells you                             |
| --------- | ----------------------------------------------- | --------------------------------------------- |
| Time      | Horizontal axis, left to right                  | When an event starts, stops or repeats        |
| Frequency | Vertical axis, low at bottom and high at top    | Pitch region and harmonic structure           |
| Intensity | Brightness or color against the color-map ruler | Relative amplitude at that time and frequency |

In the common RX color maps, black or very dark areas represent little energy and brighter areas represent more. Do not treat bright orange as “bad audio.” A clean vowel can be bright because its fundamental and harmonics are strong. A quiet electrical tone may be visually thin yet distracting.

The official [iZotope spectrogram guide](https://www.izotope.com/en/learn/understanding-spectrograms.html?ref=izotoperx.net) uses the same model: time on the horizontal axis, frequency in hertz on the vertical axis, and amplitude as brightness. Start there before memorizing artifact shapes.

## Waveform vs spectrogram vs spectrum

A waveform plots amplitude over time. It quickly reveals edits, silence, transients and flat-topped clipping, but it does not separate overlapping frequency components into bands. A spectrogram adds frequency to the time view and represents amplitude with color. The [RX Spectrum Analyzer](https://docs.izotope.com/rx12/en/spectrum-analyzer.html?ref=izotoperx.net) shows the momentary spectrum around the playhead, a selected range’s average spectrum, or the spectrum during playback. It does not lay out the full recording’s time history like the spectrogram.

RX’s [Spectrogram/Waveform display manual](https://docs.izotope.com/rx12/en/spectrogram-waveform-display.html?ref=izotoperx.net) describes a Transparency Balance slider that superimposes the two views. Move toward waveform when checking peak shape or a cut. Move toward spectrogram when locating a whistle, hiss band or cough. A middle blend is often best for clicks because you can see both the narrow time event and its broadband frequency spread.

Clipping is the important exception to “look at the spectrogram.” The current [RX 12 problem-identification guide](https://docs.izotope.com/rx12/en/identifying-audio-problems.html?ref=izotoperx.net) says the spectrogram is not particularly useful for detecting clipped audio; zoom into the waveform and look for truncated, squared-off peaks. The [De-clip guide](https://izotoperx.net/fix-clipped-audio-rx-de-clip/) shows how to confirm the threshold before processing.

## Set the display before diagnosing the audio

A spectrogram can hide a problem because of its display settings without changing the audio itself. Open **View → Spectrogram Settings**, or right-click the spectrogram and choose **Spectrogram Settings**, for display type and FFT controls. First zoom to the event. Then adjust the color-map ruler so quiet details become visible. RX lets you drag or scroll the color-map range; this reveals low-level material without applying gain to the file.

Choose a frequency scale that fits the question. Linear spaces equal intervals in hertz equally on screen. Compared with logarithmic spacing, it gives the higher region more room; logarithmic scales emphasize the low end. Mel—the RX default—tracks perceived pitch, while Bark relates to critical bands and loudness perception. If 50/60 Hz hum and its harmonics are cramped at the bottom, try a scale that gives the low region more space rather than assuming the lines are absent.

Use Auto-adjustable or Multi-resolution spectrogram types when you do not yet know which manual FFT setting you need. Adaptively Sparse can expose fine detail but is slower to calculate. High-quality rendering preserves clarity; RX may show a faster, less accurate preview across a long visible duration, then recalculate accurately when you zoom in.

## FFT size trades time detail for frequency detail

A larger FFT size improves frequency resolution: notes, hum lines and tonal events look more distinct. The tradeoff is poorer time definition, so clicks and short transients smear horizontally. A smaller FFT size sharpens timing but spreads tonal information vertically.

That tradeoff is not an audio-quality setting and does not process the file. It only changes the analysis display. For low-frequency hum, a larger FFT can separate nearby tonal lines. A plosive may also become clearer in frequency, but check a shorter FFT when you need its exact onset. For percussion, clicks and brief high-frequency events, a smaller FFT can make the timing easier to select.

Do not chase a single “best FFT size.” Zoom, change the display, then return to the setting that makes the current problem easiest to isolate. RX’s Auto-adjustable mode changes FFT size with zoom, while Multi-resolution favors frequency detail at low frequencies and time detail at high frequencies.

## A visual dictionary of common RX problems

| What you see                                        | Likely source                     | First module to inspect                         |
| --------------------------------------------------- | --------------------------------- | ----------------------------------------------- |
| Horizontal base line with evenly spaced lines above | 50/60 Hz hum and harmonics        | De-hum                                          |
| Thin horizontal line high in the display            | Whistle or higher-frequency buzz  | Spectral De-noise or targeted Spectral Repair   |
| Speckled cloud around wanted material               | Hiss, fan or HVAC broadband noise | Voice De-noise or Spectral De-noise             |
| Short bright vertical line                          | Click, pop or digital impulse     | De-click; Mouth De-click for speech-mouth noise |
| Irregular isolated shape                            | Cough, horn, bird, chair or phone | Spectral Repair                                 |
| Dark discontinuity where content should continue    | Gap or dropout                    | Spectral Repair for a short damaged region      |
| Flat or truncated waveform peaks                    | Digital clipping                  | De-clip                                         |

The RX 12 manual illustrates these patterns, but real recordings overlap. Speech produces horizontal harmonics; a click can cut vertically through them; hiss fills the background. Identify the pattern that does not belong, then listen to that exact selection. A dark patch alone does not prove lost audio. Spectral Repair can estimate a short gap from its surroundings; it cannot guarantee recovery of missing words. [Ambience Match](https://docs.izotope.com/rx12/en/ambience-match.html?ref=izotoperx.net), available in Advanced, can supply matching background ambience, not reconstruct the missing performance.

## How speech, music and noise look different

Voiced speech and pitched instruments create a fundamental with harmonics above it. On a sustained note, those appear as parallel horizontal tracks. A rising pitch curves upward. Sibilants such as S and SH spread noisy energy across the upper frequencies rather than forming neat harmonic lines. Plosives concentrate brief energy lower in the display.

Music adds many fundamentals, harmonics and transients at once. A drum hit produces a strong vertical onset followed by decay; bass occupies the lower region; cymbals create dense high-frequency texture. This is why a busy mix can look “noisy” while being healthy. Compare the suspicious event with nearby, similar musical moments before selecting it.

Steady broadband noise occupies many frequencies, often as a speckled bed behind wanted material. A steady tonal noise stays near one or several frequencies; other tones can drift. Changing traffic or wind may shift and swell. The [Voice De-noise vs Spectral De-noise comparison](https://izotoperx.net/voice-denoise-vs-spectral-denoise/) turns those visual and audible differences into a module decision.

## Choose the selection tool from the shape

The [RX 12 Interactive Tools manual](https://docs.izotope.com/rx12/en/interactive-tools.html?ref=izotoperx.net) defines the selection tools by geometry:

- **Time \[T\]:** selects every frequency across a horizontal time range.
- **Time-Frequency \[R\]:** draws a rectangle limited by time and frequency.
- **Frequency \[F\]:** selects a frequency band across time.
- **Lasso \[L\]:** outlines an irregular event.
- **Brush \[B\]:** paints a free-form region with adjustable size.
- **Magic Wand \[W\]:** click once to select the most prominent tone under the cursor; click an existing selection to include related harmonics.
- **Harmonic Selection:** extends a fundamental selection to harmonics above it.

Match the tool to the contaminant, not to habit. A click may need a narrow Time selection. A whistle can suit Frequency or Magic Wand. A cough under dialogue needs a shaped Lasso or Brush selection that avoids as much wanted speech as possible. With Instant Process off, hold Shift while drawing to add another selection, or Alt/Option to subtract from an existing one. Enable Selection Feathering with its feather icon or Shift+F; open its controls with Alt/Option+F to adjust the transition between processed and untouched audio. Higher feathering values produce longer crossfades, so check that the transition does not spread into a neighboring sound you need to preserve.

## A seven-step spectrogram reading method

1. **Hear the problem first.** Keep the original file separate, open a working copy and turn Instant Process off. Loop a short passage and describe the artifact without looking.
2. **Place it in time.** Use the waveform and playhead to mark when the sound begins and ends.
3. **Locate its frequency range.** Read bottom-to-top and compare the event with nearby wanted material.
4. **Adjust the display, not the audio.** Change zoom, transparency, color range, scale or FFT size until the structure is legible.
5. **Select by shape.** Choose Time, Time-Frequency, Frequency, Lasso, Brush, Magic Wand or Harmonic Selection.
6. **Audition the selected frequencies.** Click Play Selection Only in the transport to hear just the selected time and frequency region. Then use normal Play to hear the event in context; isolated playback can hide damage to the surrounding sound.
7. **Test one repair and keep a way back.** Preview the chosen module or use Compare where available. If a module requires rendering for audition, render only the small test selection on your working copy. Preview and Compare do not commit an edit: when satisfied, apply the chosen settings with Render. If you already rendered a test, do not render it a second time. Listen in context and Undo if useful sound is lost; export an accepted result to a new file and play that file to check it.

The [RX transport manual](https://docs.izotope.com/rx12/en/transport-controls---displays.html?ref=izotoperx.net) distinguishes Play Selection Only from normal Play. The [RX 12 beginner tutorial](https://izotoperx.net/izotope-rx-12-tutorial-beginners/) explains Preview, Compare and History. Reading the picture is only the diagnostic half; a reversible audition is what turns it into a safe edit.

## Three mistakes that make the display lie to you

**Judging a long-file preview as final detail.** RX can lower display accuracy when a long duration is visible. Zoom into the event before concluding a click or narrow tone is absent.

**Changing the color range until everything looks bright.** The color map is relative analysis, not a cleanliness score. Revealing quiet room tone is useful; treating every revealed speckle as damage leads to overprocessing.

**Selecting by appearance without playback.** A harmonic that resembles hum may belong to a bass note. A vertical line may be an intentional snare transient. A dark gap may be edited silence. Sound confirms meaning.

For illustrated examples, iZotope’s spectrogram tutorial linked above compares speech, noise and transient shapes. Some screenshots there are labeled RX 11, so use them to understand the patterns and the RX 12 manual for current controls. The picture should help you ask a better listening question, not make the decision for you.

## Turn the shape into the right RX module

The spectrogram should narrow the tool choice. Horizontal electrical ladders point toward De-hum. A continuous speckled floor points toward a denoiser. Vertical impulses point toward De-click. Isolated irregular events point toward Spectral Repair. Flat waveform peaks point toward De-clip. Room decay after each word suggests De-reverb or Dialogue Isolate rather than a frequency selection.

Do not apply Spectral Repair to every visible anomaly. Dedicated modules can be more practical for repeated clicks or steady noise than outlining every event by hand. Check the chosen module’s selection behavior rather than assuming every processor respects a frequency-shaped selection. Use the [Repair Audio hub](https://izotoperx.net/repair-audio-rx/) to move from artifact class to the specialized guide, or the [Clean Dialogue hub](https://izotoperx.net/clean-dialogue-rx/) when the wanted signal is speech.

## When a spectrogram is not enough

A spectrogram does not tell you whether a repair sounds natural in context, whether a noise will be masked by music, or whether a delivery meets loudness requirements. It also cannot recover information that was never captured. Treat it as a map: it shows where energy lives, not what every pixel means.

Work in short cycles—hear, see, select, preview, compare. Preserve the original file separately; an Undo History state is not a substitute for a disk backup, especially if you overwrite and close the audio file. If the visual result looks clean but the voice sounds phasey or thin, trust the sound and undo. If the repair sounds right while a faint visual trace remains, the trace may be the better compromise.

For the broader learning path, return to the [Learn iZotope RX hub](https://izotoperx.net/learn-izotope-rx/). The spectrogram is the navigation skill that makes every later module tutorial easier.

## Frequently asked questions

### What do the axes show in the RX spectrogram?

The horizontal axis shows time, the vertical axis shows frequency with low frequencies at the bottom, and color or brightness shows relative intensity.

### What does a horizontal line mean in a spectrogram?

It usually indicates a sustained tonal component. A base line at 50 or 60 Hz with evenly spaced lines above often points to electrical hum and its harmonics.

### What does a vertical line mean in the RX spectrogram?

A short vertical line usually indicates a transient that spans many frequencies, such as a click, pop or sharp edit. It can also be a wanted drum attack or consonant. Confirm it by playing a narrow time selection.

### Why does hiss look like speckles?

Hiss is broadband noise spread across many frequencies rather than one stable tone, so it appears as a continuous speckled field around wanted material.

### Should I use a large or small FFT size?

Use a larger FFT when tonal and low-frequency detail matters; use a smaller FFT when the timing of short transients matters. Auto-adjustable or Multi-resolution modes are practical starting points.

### Can I identify clipping from the spectrogram?

Use the waveform instead. Zoom in and look for genuinely truncated, squared-off peaks; dense or limited audio can look flat when zoomed out without proving clipping.

### Which RX selection tool should beginners use first?

Turn Instant Process off, then start with Time-Frequency \[R\] for a rectangular event. Use Lasso or Brush for irregular shapes and Magic Wand or Harmonic Selection for tonal structures. Check the region with Play Selection Only.

### Does changing spectrogram settings change the audio?

No. Zoom, color map, frequency scale, FFT size and display type change the visualization, not the underlying file. Processing and editing do change audio, including drawing a new selection while Instant Process is enabled. Keep that modifier off while exploring the display.