--- title: "Audio EQ" description: "Equalize a voice clip: a presence bell (cut harshness in the 2-5 kHz band), a de-esser (tames sibilance around 6-8 kHz, only while it is loud), and a low shelf (body). The default preset `voice_deharsh` is the fix for harsh TTS clones (Fish, ElevenLabs): -3.5 dB wide bell at 3.5 kHz, de-ess at 7 kHz, +1.5 dB shelf at 150 Hz. `none` starts every band off. Any dial you set overrides the preset for that one value; a band with gain 0 (de-ess amount 0) is off. When bands boost, the whole clip is first turned down by the sum of the boosts, so the EQ cannot push it past full scale; follow with a loudness node if level matters. Out-of-range values are refused with the number, never clamped. Sample rate and channels are kept. Free: deterministic processing, content-addressed output." updated: 2026-09-28 --- # Audio EQ **Node type:** `utility:audio_eq` **Category:** `Editing` ## Description Equalize a voice clip: a presence bell (cut harshness in the 2-5 kHz band), a de-esser (tames sibilance around 6-8 kHz, only while it is loud), and a low shelf (body). The default preset `voice_deharsh` is the fix for harsh TTS clones (Fish, ElevenLabs): -3.5 dB wide bell at 3.5 kHz, de-ess at 7 kHz, +1.5 dB shelf at 150 Hz. `none` starts every band off. Any dial you set overrides the preset for that one value; a band with gain 0 (de-ess amount 0) is off. When bands boost, the whole clip is first turned down by the sum of the boosts, so the EQ cannot push it past full scale; follow with a loudness node if level matters. Out-of-range values are refused with the number, never clamped. Sample rate and channels are kept. Free: deterministic processing, content-addressed output. ## Canvas ports These appear as port handles on the left side of the node. | ID | Label | Details | | ------- | --------- | -------------------- | | `audio` | **Audio** | `AUDIO` _(required)_ | ## Sidebar config These render as form fields in the right-side config panel when the node is selected. | ID | Label | Details | | -------------- | ------------------------------------------------ | ----------------------------------------- | | `preset` | **Preset** | `TEXT` · options: `voice_deharsh`, `none` | | `presenceGain` | **Presence gain (dB, -12 to 6; blank = preset)** | `NUMBER` | | `presenceFreq` | **Presence frequency (Hz, 1000 to 8000)** | `NUMBER` | | `presenceQ` | **Presence Q (0.3 to 4; lower = wider)** | `NUMBER` | | `deessAmount` | **De-ess amount (0 to 1; 0 = off)** | `NUMBER` | | `deessFreq` | **De-ess frequency (Hz, 4000 to 10000)** | `NUMBER` | | `lowShelfGain` | **Low shelf gain (dB, -6 to 6)** | `NUMBER` | | `lowShelfFreq` | **Low shelf frequency (Hz, 50 to 400)** | `NUMBER` | | `format` | **Format** | `TEXT` · options: `mp3`, `wav` | ## Outputs | ID | Label | Type | | ------- | ------------ | ------- | | `audio` | **EQ Audio** | `AUDIO` | ## What it is for TTS voice clones (Fish, ElevenLabs) often come out harsh: the 2-5 kHz band sits 7 to 10 dB under the low end, where a natural voice sits about 15 dB under, and the S sounds spit. Wire the voice into Audio EQ and the default preset fixes both without a trip to an audio editor. ## The voice_deharsh preset - **Presence**: -3.5 dB, a wide bell (Q 0.7) centred at 3.5 kHz. - **De-ess**: a dynamic cut around 7 kHz that only engages while that band is loud, so it takes the edge off sibilants and leaves everything else alone. - **Low shelf**: +1.5 dB at 150 Hz, a little body back. Measured on real Fish clones, the preset moves the 2-5 kHz band 3 to 3.5 dB further under the low end and trims the 6-8 kHz sibilance band by 2 to 3 dB. Want more? Set **Presence gain** to -6. ## Settings Every dial starts blank, which means "use the preset's value". Set one and it overrides that value only. Pick the **none** preset to start from flat. - **Presence gain** (dB, -12 to 6) and **frequency** (1000 to 8000 Hz), **Q** (0.3 to 4, lower is wider). Gain 0 (anything under 0.05 dB) turns the band off. - **De-ess amount** (0 to 1; under 0.01 is off) and **frequency** (4000 to 10000 Hz). - **Low shelf gain** (dB, -6 to 6) and **frequency** (50 to 400 Hz). - **Format**: mp3 (default) or wav. Sample rate and channels are kept. Every band off is refused rather than handing back a transcode. Out-of-range values are refused with the number, never clamped. ## Headroom When bands boost, the whole clip is first turned down by the SUM of the boosts, so a voice already normalized near full scale does not clip. The sum, because two boosts that overlap (a +6 shelf and a wide +6 bell) add up past either one. The preset boosts 1.5 dB, so its output is about 1.5 dB quieter at peak. Put **Audio Master** (or your loudness step) after it if the level matters. ## Free, and why Free set, no Run button, refreshes when anything downstream runs. The output is content-addressed by the input url and the settings, so an unchanged re-run reuses the last file. --- _Auto-generated from the Wireflow node registry._ --- Documentation index: fetch https://www.wireflow.ai/llms.txt for the full list of Wireflow docs.