Back to Details

davinci-resolve-audio-cleanup-podcast

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

DaVinci Resolve — clean podcast and interview audio

DaVinci Resolve — 播客与访谈音频清理

Overview

概述

The four-step podcast audio cleanup chain: EQ out rumble -> kill any hum -> denoise -> level -> duck music. Done in this order. The principle: address frequency problems before dynamic problems. EQ is surgical; noise reduction is destructive. Use the surgical tool first so the destructive tool has less to do.
播客音频清理的四步流程：EQ去除低频杂音 → 消除嗡嗡声 → 降噪 → 平衡音量 → 压低背景音乐。需按此顺序操作。原则：先解决频率问题，再处理动态问题。EQ是精准调整工具；降噪具有破坏性。先使用精准工具，减少破坏性工具的处理量。

When to use

适用场景

Symptoms:
  • Recording has background hum (electrical 50/60 Hz interference)
  • Air conditioning, refrigerator, or traffic noise floor under dialogue
  • Voice levels vary from clip to clip (loud intros, quiet thoughtful moments)
  • Music score drowns out the host when they start talking
  • Sibilance ("ess" sounds) too harsh
When NOT to use:
  • Multi-mic interview where you need to gate each mic separately — that needs the Track-level Dynamics workflow per mic (Fairlight Audio Post pp. 373-375).
  • Audio is destroyed (clipped, severe wind noise, water damage) — there is a limit to what cleanup can do. Re-record if possible.
  • You need surround/Atmos mixing — different chapter (Fairlight Audio Post pp. 643+).
适用症状：
  • 录音存在背景嗡嗡声（50/60Hz电气干扰）
  • 对话下方存在空调、冰箱或交通噪音
  • 不同片段的语音音量差异大（开头响亮，思考时声音微弱）
  • 背景音乐盖过主持人说话声
  • 齿音（“嘶”声）过于刺耳
不适用场景：
  • 需要单独对每个麦克风设置门限的多麦克风访谈——这种情况需要针对每个麦克风使用轨道级动态处理流程（参考《Fairlight Audio Post》第373-375页）。
  • 音频已损坏（削波、严重风噪、进水损坏）——清理能力有限，如有可能建议重新录制。
  • 需要环绕声/Atmos混音——属于另一章节内容（参考《Fairlight Audio Post》第643页及以后）。

Quick reference — the cleanup chain in order

快速参考——清理流程顺序

Phase -1 — source audit before cleanup

准备阶段——清理前的源文件检查

Before applying effects, audit every candidate audio source. Do not assume the named lav/boom file is best. Check:
  • duration matches picture
  • silence percentage
  • integrated LUFS
  • true peak
  • clipping risk
  • audible room tone/hum
  • whether the track is actually silent
For multi-camera workshop folders, keep the selected dialogue on 
A1
, the processed dialogue on 
A2
, music/ambience on 
A3
, and SFX/bumpers on 
A4
. If the external mic peaks near 
0 dBFS
, preserve it as source but create a processed duplicate with conservative limiting/leveling before delivery.
StepToolWhen to useWhere
1. EQ rumbleClip EQ (Inspector > Equalizer)Always — every dialogue clip has sub-100Hz garbageInspector panel, audio clip selected
2. De-HummerFairlight FX > De-HummerPower line buzz (50 Hz Europe, 60 Hz US)Drag from Effects Library onto clip
3. Gate the noise floorMixer > Dynamics > GateAir conditioner / room tone audible between sentencesMixer track dynamics panel
4. Noise ReductionFairlight FX > Noise ReductionStubborn broadband noise that survives steps 1-3Drag from Effects Library onto clip or track
5. AI Dialogue LevelerInspector > AI Dialogue Leveler (Studio) or manual keyframesVoice gets loud/quiet within a sentenceInspector Track FX section
6. DuckerTrack FX > DuckerMusic too loud under dialogueInspector on the music track
在添加效果前，检查所有候选音频源。不要默认认为命名为领夹麦/吊杆麦的文件就是最佳选择。检查：
  • 时长是否与画面匹配
  • 静音占比
  • 综合LUFS值
  • 真实峰值
  • 削波风险
  • 可听见的房间混响/嗡嗡声
  • 轨道是否真的静音
对于多机位工作文件夹，将选中的对话音频保留在
A1
轨道，处理后的对话音频放在
A2
，音乐/环境音放在
A3
，音效/过渡音放在
A4
。如果外接麦克风的峰值接近
0 dBFS
，保留源文件，但在交付前创建一个经过保守限制/平衡处理的副本。
步骤工具适用场景位置
1. EQ去除低频杂音Clip EQ（检查器 > 均衡器）所有场景——每个对话片段都存在100Hz以下的无用杂音检查器面板，选中音频片段
2. 去嗡嗡声Fairlight FX > De-Hummer电源线嗡嗡声（欧洲50Hz，美国60Hz）从效果库拖到片段上
3. 噪音底部门限混音器 > 动态 > 门限句子间隙可听见空调/房间噪音混音器轨道动态面板
4. 降噪Fairlight FX > Noise Reduction步骤1-3无法消除的顽固宽带噪音从效果库拖到片段或轨道上
5. AI对话音量平衡器检查器 > AI Dialogue Leveler（Studio版本）或手动关键帧句子内语音音量忽大忽小检查器轨道FX区域
6. 音量压低器轨道FX > Ducker背景音乐盖过对话音乐轨道的检查器

Steps

详细步骤

Phase 0 — Switch to Fairlight, set up your monitoring

准备阶段——切换到Fairlight页面，设置监听

  1. Press Shift-7 to switch to the Fairlight page.
  2. Open the Mixer (Toolbar > Mixer button).
  3. Put on headphones. You will hear problems on headphones that you will miss on laptop speakers.
  1. 按 Shift-7 切换到Fairlight页面。
  2. 打开混音器（工具栏 > 混音器按钮）。
  3. 戴上耳机。耳机能听到笔记本扬声器无法察觉的问题。

Step 1 — Clip EQ: cut everything below 80 Hz (always)

步骤1——Clip EQ：切除80Hz以下所有频率（必做）

This is non-negotiable. Every voice recording has rumble, HVAC, traffic, and breath plosives under 80 Hz that contribute nothing to intelligibility but eat up headroom.
  1. Select a dialogue clip in the timeline.
  2. Open the Inspector (top right of the interface, then Audio tab).
  3. Scroll down to Equalizer. Turn it ON (the radio button next to "Equalizer").
  4. By default, Bands 2-5 are enabled. Set Band 2 to Low Shelf (the first icon in the Band Filter Type dropdown).
  5. Drag the Band 2 handle in the graph down to -70 dB and right until the shelf reaches the 80-100 Hz line.
  6. Repeat for every dialogue clip — or copy attributes from this clip to all others (Cmd-C on the clip, select others, Edit > Paste Attributes, check Equalizer only, Apply).
Fairlight Audio Post pp. 368-372 walks through this with the same EQ controls.
这一步必不可少。所有语音录制都存在低频杂音、空调声、交通声和80Hz以下的爆破音，这些对清晰度毫无帮助，还会占用动态余量。
  1. 在时间轴中选中一个对话片段。
  2. 打开检查器（界面右上角，然后切换到音频标签）。
  3. 向下滚动到均衡器。开启均衡器（“Equalizer”旁的单选按钮）。
  4. 默认情况下，频段2-5已启用。将频段2设置为低架滤波器（频段滤波器类型下拉菜单中的第一个图标）。
  5. 将频段2的控制点拖到**-70 dB**，并向右拖动直到滤波器覆盖80-100Hz范围。
  6. 对所有对话片段重复此操作——或者复制属性（选中该片段按Cmd-C，选中其他片段，编辑 > 粘贴属性，仅勾选均衡器，应用）。
《Fairlight Audio Post》第368-372页详细介绍了相同的EQ操作。

Step 2 — De-Hummer for any electrical hum

步骤2——使用De-Hummer消除电气嗡嗡声

If you hear a constant tone behind the voice — usually a low buzz at 50 Hz or 60 Hz from poor cable shielding or a ground loop — use the De-Hummer.
  1. Open the Effects Library (top-left button on Fairlight page).
  2. Under Audio FX > Fairlight FX, find De-Hummer.
  3. Drag De-Hummer onto the affected clip in the timeline.
  4. In the De-Hummer window, click 60 Hz (or 50 Hz if you are in Europe / Asia / Africa).
  5. Drag the Amount knob right until the hum disappears. Typical setting: 
    -20 to -30 dB
    .
  6. Adjust the Slope knob slightly right if you still hear harmonics (multiples of 60 Hz — 120, 180, 240).
  7. Click the Bypass switch at the top to A/B compare. The Bypass switch is red when the filter is on.
Fairlight Audio Post pp. 376-380 covers this exactly. The trick is that the De-Hummer is a notch filter — it removes specific frequencies without affecting nearby ones, unlike EQ.
如果听到语音背后有持续的音调——通常是50Hz或60Hz的低频嗡嗡声，由线缆屏蔽不良或接地回路导致——使用De-Hummer。
  1. 打开效果库（Fairlight页面左上角按钮）。
  2. 在音频FX > Fairlight FX下找到De-Hummer。
  3. 将De-Hummer拖到时间轴中受影响的片段上。
  4. 在De-Hummer窗口中，点击60 Hz（如果在欧洲/亚洲/非洲则点击50 Hz）。
  5. 将Amount旋钮向右拖动直到嗡嗡声消失。典型设置：
    -20至-30 dB
    。
  6. 如果仍能听到谐波（60Hz的倍数——120、180、240Hz），稍微向右调整Slope旋钮。
  7. 点击顶部的Bypass开关进行A/B对比。开关为红色时表示滤镜已启用。
《Fairlight Audio Post》第376-380页对此有详细说明。关键在于De-Hummer是一个陷波滤波器——它能移除特定频率而不影响附近频率，这与EQ不同。

Step 3 — Gate the noise floor (track-level dynamics)

步骤3——噪音底部门限（轨道级动态处理）

If there is a low-level hiss, computer fan, or AC noise that is audible between sentences but masked when the voice is playing, a gate kills it.
  1. In the Mixer, find the dialogue track's channel strip.
  2. Double-click the Dynamics section to open the Dynamics panel.
  3. Click the Gate button to enable.
  4. Start with these settings:
    • Threshold: -35 dB (default)
    • Range: -18 dB (start conservative)
    • Attack: 1.4 ms
    • Hold: 0 ms
    • Release: 93 ms
  5. Play back the track. The Gain Reduction meter should activate only between sentences (when no voice). If the gate cuts off the start of words, raise the threshold (less negative). If it cuts off breaths or quiet sentence endings, lower the range.
The exact starting values come from Fairlight Audio Post pp. 373-375.
如果存在低电平嘶嘶声、电脑风扇声或空调声，且在句子间隙可听见，但说话时被掩盖，门限工具可以消除它。
  1. 在混音器中找到对话轨道的通道条。
  2. 双击动态区域打开动态面板。
  3. 点击Gate按钮启用。
  4. 从以下设置开始：
    • Threshold：-35 dB（默认值）
    • Range：-18 dB（初始保守设置）
    • Attack：1.4 ms
    • Hold：0 ms
    • Release：93 ms
  5. 播放轨道。增益衰减表应仅在句子间隙（无语音时）激活。如果门限切断了单词的开头，提高阈值（减小负值）。如果切断了呼吸声或微弱的句尾，降低Range值。
精确的初始值来自《Fairlight Audio Post》第373-375页。

Step 4 — Noise Reduction (when steps 1-3 are not enough)

步骤4——降噪（步骤1-3不足以解决时）

Some recordings have broadband noise (general "hisssss") that EQ cannot remove and the gate would chop up too aggressively. Use the Fairlight FX Noise Reduction.
  1. From the Effects Library > Audio FX > Fairlight FX, drag Noise Reduction onto the affected clip.
  2. The Noise Reduction panel opens. You have two modes:
    • Manual (default) — you "teach" the plug-in what the noise sounds like
    • Auto Speech Mode — algorithm separates speech from noise automatically
  3. For Auto Speech Mode: just enable it. Adjust the Dry/Wet slider (default 80%). Lower it if voices start sounding "underwater."
  4. For Manual mode: a. Find a section of the clip with only noise (silent gap, ideally at the start before anyone speaks). b. Click Learn to enable noise-profile capture. c. Play the noise-only section. d. Click Learn again to stop. The plug-in has now "learned" the noise. e. Play back the full clip — noise should drop dramatically.
Fairlight Audio Post pp. 381-383 covers both modes. Studio version only — free Resolve does not have the Noise Reduction plug-in. Free users: use a more aggressive Clip EQ (step 1) combined with the Gate (step 3).
有些录音存在宽带噪音（普遍的“嘶嘶声”），EQ无法移除，门限处理又会过度切割音频。此时使用Fairlight FX Noise Reduction。
  1. 从效果库 > 音频FX > Fairlight FX中，将Noise Reduction拖到受影响的片段上。
  2. 打开Noise Reduction面板。有两种模式：
    • 手动（默认）——你“教”插件识别噪音的特征
    • Auto Speech Mode——算法自动分离语音和噪音
  3. 自动语音模式操作：只需启用它。调整Dry/Wet滑块（默认80%）。如果语音听起来像“在水下”，降低该值。
  4. 手动模式操作： a. 找到片段中只有噪音的部分（静音间隙，理想情况是在有人说话前的开头）。 b. 点击Learn启用噪音特征捕捉。 c. 播放只有噪音的部分。 d. 再次点击Learn停止。插件现在已“学习”了噪音特征。 e. 播放整个片段——噪音应大幅降低。
《Fairlight Audio Post》第381-383页介绍了两种模式。仅Studio版本可用——免费版Resolve没有Noise Reduction插件。免费版用户：使用更激进的Clip EQ（步骤1）结合门限工具（步骤3）。

Step 5 — AI Dialogue Leveler (Studio) or manual keyframes (Free)

步骤5——AI对话音量平衡器（Studio版本）或手动关键帧（免费版）

Voices have natural dynamic range — a person gets quieter as they finish a sentence. The leveler evens this out.
Studio approach (one click):
  1. Select the dialogue clip.
  2. In the Inspector > Track FX section, enable AI Dialogue Leveler.
  3. That is it. Adjust the strength slider if it sounds too "compressed."
Free Resolve approach (manual keyframes):
  1. In the Edit page, Option-click (Mac) / Alt-click (Windows) the volume bar in an audio clip to add keyframes.
  2. Add a keyframe before a quiet section, another keyframe after.
  3. Drag the volume bar between keyframes up (typically +3 to +6 dB) to boost the quiet section.
Beginner's Guide pp. 150-153 covers the manual keyframe workflow with screenshots.
语音有自然的动态范围——人在说完句子时声音会变小。平衡器可以消除这种差异。
Studio版本操作（一键完成）：
  1. 选中对话片段。
  2. 在检查器 > 轨道FX区域，启用AI Dialogue Leveler。
  3. 完成。如果声音听起来过于“压缩”，调整强度滑块。
免费版Resolve操作（手动关键帧）：
  1. 在编辑页面，按住Option键（Mac）/Alt键（Windows）点击音频片段的音量条添加关键帧。
  2. 在安静部分前后各添加一个关键帧。
  3. 拖动两个关键帧之间的音量条向上（通常+3至+6 dB）以提升安静部分的音量。
《Beginner's Guide》第150-153页配有截图介绍手动关键帧流程。

Step 6 — Duck the music under dialogue

步骤6——压低背景音乐以突出对话

If you have music underneath voice and the music swallows the voice, use the Ducker.
  1. Select the MUSIC track (or the music clip).
  2. In the Inspector > Track FX section, enable Ducker.
  3. In the Ducker controls:
    • Source 1 dropdown: set to your dialogue track (e.g. DIALOGUE).
    • Duck Level: start at 2.7 dB (the default — a subtle duck).
    • If you have multiple dialogue tracks (host + guest), click the + to add Source 2 and Source 3.
  4. Play back. Watch the Ducker Graph — when dialogue plays, the music level dips automatically.
  5. If the duck is too dramatic, lower Duck Level. If you cannot hear the voice over music, raise Duck Level.
Fairlight Audio Post pp. 552-558 covers Ducker in depth, including the warning that Duck Level > 5 dB sounds amateurish because the audience perceives the dip.
如果背景音乐盖过了语音，使用Ducker。
  1. 选中音乐轨道（或音乐片段）。
  2. 在检查器 > 轨道FX区域，启用Ducker。
  3. 在Ducker控制中：
    • Source 1下拉菜单：设置为你的对话轨道（例如DIALOGUE）。
    • Duck Level：从2.7 dB开始（默认值——轻微压低）。
    • 如果有多个对话轨道（主持人+嘉宾），点击**+**添加Source 2和Source 3。
  4. 播放。观察Ducker图表——当对话播放时，音乐音量会自动降低。
  5. 如果压低效果过于明显，降低Duck Level。如果仍听不到语音，提高Duck Level。
《Fairlight Audio Post》第552-558页详细介绍了Ducker，包括警告：Duck Level超过5 dB会显得不专业，因为观众会察觉到音量变化。

Common mistakes

常见错误

  • Applying Noise Reduction first, then EQ -> wrong order. EQ out the obvious frequencies first so the Noise Reduction has less to remove and is less destructive. (This is the misconception this skill addresses head-on.) Fairlight Audio Post p. 368-372 establishes EQ before p. 381-383 noise reduction for this reason.
  • Setting the Gate threshold too high -> it cuts off the start of each word. Listen carefully and back the threshold off.
  • Cranking the Ducker Duck Level above 5 dB -> the audience hears the level swing and it feels artificial. Stay subtle.
  • Forgetting that the Bypass switch is RED when ACTIVE -> the convention is inverted from what you might expect. Red = filter is engaged.
  • Trying to fix wind noise -> the Fairlight FX cannot meaningfully fix wind. Re-record or use iZotope RX externally.
  • 先应用降噪，再用EQ -> 顺序错误。先通过EQ去除明显的频率问题，这样降噪工具需要处理的内容更少，破坏性也更小。（这是本技能重点纠正的误解。）《Fairlight Audio Post》第368-372页先介绍EQ，再在第381-383页介绍降噪，正是出于这个原因。
  • 门限阈值设置过高 -> 会切断每个单词的开头。仔细聆听并降低阈值。
  • Ducker的Duck Level设置超过5 dB -> 观众会察觉到音量波动，显得不自然。保持效果微妙。
  • 忘记Bypass开关为红色时表示已启用 -> 这个惯例与预期相反。红色=滤镜已激活。
  • 尝试修复风噪 -> Fairlight FX无法有效修复风噪。重新录制或使用外部工具iZotope RX。

Verification

验证标准

You succeeded if all of the following are true:
  1. Listening on headphones at moderate volume, you cannot hear hum, hiss, or AC noise between sentences.
  2. The loudness meter (top right of Fairlight) reads between -16 and -20 LUFS over the entire timeline (this is podcast-standard loudness).
  3. Voice clips peak at around -9 dBFS to -3 dBFS but never clip (red).
  4. Music plays at full volume when there is no voice, ducks audibly but smoothly when voice starts, returns to full when voice ends.
  5. Bypassing the Clip EQ on the dialogue clip makes the audio sound noticeably "boomier" and worse — the EQ is doing something audible.
满足以下所有条件即为成功：
  1. 用中等音量的耳机聆听，句子间隙听不到嗡嗡声、嘶嘶声或空调噪音。
  2. Fairlight右上角的响度计显示整个时间轴的响度在-16至-20 LUFS之间（这是播客标准响度）。
  3. 语音片段的峰值在-9 dBFS至-3 dBFS之间，且从未出现削波（红色）。
  4. 无语音时音乐以全音量播放，语音开始时音乐音量明显但平滑地降低，语音结束后恢复全音量。
  5. 关闭对话片段的Clip EQ后，音频听起来明显“更浑浊”且更差——说明EQ起到了作用。

Transfer

拓展应用

Now try this on a 3-person podcast where each person has their own mic (so three separate dialogue tracks). The cleanup chain is the same per track, but the Ducker setup changes: in the music track's Ducker, add Source 1, 2, and 3 pointing at each of the three dialogue tracks. The music will duck when any of the three is speaking. If two people talk over each other, the duck does not stack — you get the same -2.7 dB dip whether one or three voices are active.
现在尝试将此流程应用于三人播客，每个人都有独立麦克风（三个独立对话轨道）。每个轨道的清理流程相同，但Ducker设置有所变化：在音乐轨道的Ducker中，添加Source 1、2、3分别指向三个对话轨道。当任意一个人说话时，音乐都会压低。如果两人同时说话，压低效果不会叠加——无论是一个还是三个人说话，都是-2.7 dB的压低幅度。

Working reference

参考文档

  • docs/wiki/fairlight-audio-post.md#lesson-6--audio-repairs-pp-367-415
    (full cleanup chain — primary)
  • docs/wiki/fairlight-audio-post.md#lesson-9--mixing-and-sweetening-pp-535-612
    (Ducker setup)
  • docs/wiki/beginners-guide.md#lesson-3--audio-editing-and-quick-export-pp-141-179
    (free Resolve fallback — manual keyframes + Fairlight FX)
  • docs/wiki/editors-guide.md#lesson-8--audio-editing-pp-475-543
    (AI Dialogue Leveler reference)
  • docs/wiki/master.md#reset-matrix--when-the-user-pushes-back-read-this
    (audio symptom rows — hum, hiss, ducker)
  • docs/wiki/fairlight-audio-post.md#lesson-6--audio-repairs-pp-367-415
    （完整清理流程——主要参考）
  • docs/wiki/fairlight-audio-post.md#lesson-9--mixing-and-sweetening-pp-535-612
    （Ducker设置）
  • docs/wiki/beginners-guide.md#lesson-3--audio-editing-and-quick-export-pp-141-179
    （免费版Resolve替代方案——手动关键帧+Fairlight FX）
  • docs/wiki/editors-guide.md#lesson-8--audio-editing-pp-475-543
    （AI Dialogue Leveler参考）
  • docs/wiki/master.md#reset-matrix--when-the-user-pushes-back-read-this
    （音频症状列表——嗡嗡声、嘶嘶声、Ducker问题）

When the agent's work isn't matching expectations (context-rot reset)

当Agent的工作不符合预期时（重置上下文）

If the user reports remaining hum, hiss, robotic ducking, or that the chain order was different, read these PDF page ranges to reset:
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    pp. 368-372 (Clip EQ for low-frequency reduction)
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    pp. 373-375 (Dynamics Gate, gating low-level noise)
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    pp. 376-380 (De-Hummer for power line hum)
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    pp. 381-389 (Noise Reduction — manual and Auto Speech Mode)
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    pp. 552-558 (Ducker — Duck Level cap at 5 dB)
  • DaVinci-Resolve-20_Beginners-Guide.pdf
    pp. 142-158 (manual keyframes + Fairlight FX for free Resolve)
  • DaVinci-Resolve-20-Editors-Guide.pdf
    p. 514 (AI Dialogue Leveler one-click voice leveling)
如果用户反馈仍存在嗡嗡声、嘶嘶声、机械感的音量压低，或流程顺序有误，请阅读以下PDF页面范围重置认知：
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    第368-372页（Clip EQ用于低频削减）
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    第373-375页（动态门限，消除低电平噪音）
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    第376-380页（De-Hummer用于电源线嗡嗡声）
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    第381-389页（降噪——手动和自动语音模式）
  • DaVinci-Resolve-20-Fairlight-Audio-Post.pdf
    第552-558页（Ducker——Duck Level上限为5 dB）
  • DaVinci-Resolve-20_Beginners-Guide.pdf
    第142-158页（免费版Resolve的手动关键帧+Fairlight FX）
  • DaVinci-Resolve-20-Editors-Guide.pdf
    第514页（AI Dialogue Leveler一键语音平衡）