Podcast Loudness Normalizer (LUFS, True Peak) for AI Agents

The Loudness Normalization API from ENT-E turns any podcast, voice or video file URL into a broadcast-grade file at -16, -14, -23 or -24 LUFS, with a true-peak ceiling, so every episode meets platform specs automatically, with no DAW, plugins or ffmpeg setup.

Loudness normalization API: hit platform LUFS targets in one call

Send one URL, or a whole batch. Pick a preset (podcast, apple, spotify, youtube, ebu-r128, atsc-a85) or set your own target. You get back a normalized MP3, M4A, WAV, FLAC or MP4 file, plus a before/after loudness report you can log, check or show to a client.

How it works:

1. The Actor measures the whole file first (two-pass EBU R128 loudness normalization with FFmpeg's loudnorm, linear mode where possible).
2. It applies gain with a 4× oversampled true-peak limiter.
3. It re-measures the encoded file and re-renders when the ceiling or target is missed, up to 3 attempts. The row you receive reports the file that was actually delivered.

What you get

• Platform presets with fixed targets:

  preset	Integrated loudness
  podcast, apple	-16 LUFS
  spotify, youtube	-14 LUFS
  ebu-r128	-23 LUFS
  atsc-a85	-24 LUFS
  custom	your targetLufs (-70 to -5)

  The true-peak ceiling defaults to -1 dBTP (truePeakDbtp) and the loudness range to 11 LU (loudnessRange).

• Measured, not assumed. Every row includes beforeLufs, afterLufs, afterTruePeakDbtp, gainAppliedDb and truePeakCompliant, plus full before/after objects with integrated loudness, loudness range and true peak.

• Output formats:

  • mp3 (192 kbps), the default
  • m4a (AAC 192 kbps)
  • wav (24-bit)
  • flac
  • mp4, which copies the video stream and re-encodes the audio to AAC, so you can normalize a video podcast without re-rendering the picture

• Batch. Use files for many episodes in one run. Any setting can be overridden per file.

• Failed files are free. A bad URL, a silent file, an audio-free input or an over-length file returns an error row and isn't charged.

Read this first: true peak beats the target

The true-peak ceiling always takes priority over the loudness target. For loud, very peaky audio, reaching -14 LUFS on the spotify or youtube preset can be impossible without crossing the -1 dBTP ceiling. In that case the Actor keeps the ceiling and lets the file land a little below -14 LUFS. You still get a clean, non-clipping file, and the row tells you exactly what happened: the measured afterLufs, truePeakCompliant: true, and normalizationType: "dynamic". No row will claim -14.0 LUFS if the file isn't at -14.0 LUFS. In our tests, integrated loudness typically lands within ±0.5 LU of the target.

Use cases

Normalize audio to LUFS for a podcast feed (-16 LUFS)

Apple Podcasts and most podcast hosts recommend around -16 LUFS. This is the default.

{
  "url": "https://example.com/episodes/ep-142-raw.wav",
  "preset": "podcast",
  "outputFormat": "mp3"
}

Spotify -14 LUFS and YouTube loudness

{
  "url": "https://example.com/episodes/ep-142-master.wav",
  "preset": "spotify",
  "truePeakDbtp": -1
}

For a video podcast going to YouTube, keep the picture and fix only the audio:

{
  "url": "https://example.com/video/ep-142.mp4",
  "preset": "youtube",
  "outputFormat": "mp4"
}

Apple Podcasts loudness with a stricter true-peak limiter

Some distributors ask for -2 dBTP or lower to leave headroom for lossy encoders.

{
  "url": "https://example.com/episodes/ep-142.wav",
  "preset": "apple",
  "truePeakDbtp": -2,
  "outputFormat": "m4a"
}

EBU R128 and ATSC A/85 for broadcast delivery

{
  "url": "https://example.com/broadcast/segment-07.wav",
  "preset": "ebu-r128",
  "outputFormat": "wav",
  "sampleRate": 48000
}

Use "preset": "atsc-a85" for -24 LUFS (US broadcast). For anything else, use "preset": "custom" with "targetLufs": -18.

Batch normalize audio for a whole back catalog

Top-level settings are the defaults, and each item in files can override them. Plain URL strings work too.

{
  "preset": "podcast",
  "outputFormat": "mp3",
  "channels": "mono",
  "files": [
    "https://example.com/episodes/ep-001.mp3",
    "https://example.com/episodes/ep-002.mp3",
    { "url": "https://example.com/promos/trailer.wav", "preset": "spotify", "outputFormat": "wav" }
  ]
}

Files are processed one after another in the same run. Each one gets its own dataset row.

Pricing (pay per event)

Event	Price
File normalized (per successful file)	$0.01
Audio minute (per started minute of input audio)	$0.008
Actor start (per run, one event per GB of memory, minimum one; default 4 GB = 4 events)	$0.00005 per event

There are no platform usage fees on top (compute is included). Failed files are free.

Worked example: a weekly one-hour show

• One 60-minute episode: $0.01 + 60 × $0.008 = $0.49.
• One year of episodes (52) in one batch run: 52 × $0.49 = $25.48, plus $0.0002 actor start.
• A 59 min 13 s recording counts as 60 started minutes, so it is also $0.49.
• A 30-second promo: $0.01 + 1 × $0.008 = $0.018.

Billing counts started minutes of the input file, rounded up, with a minimum of 1. Up to 0.5 s of container padding is ignored, so a 60 min 00.4 s file counts as 60 minutes, while 60 min 01 s counts as 61 ($0.498). Set a spending limit on the run to cap the total; the Actor stops cleanly when the limit is reached.

Input reference

Field	Type	Default	Notes
url	string	(none)	Direct link to one audio or video file. Any FFmpeg-readable format, up to 2 GB. The first audio stream is normalized.
files	array	(none)	Batch. Items are { "url": ..., <any setting below> } or plain URL strings.
preset	string	podcast	podcast, apple, spotify, youtube, ebu-r128, atsc-a85, custom
targetLufs	number	(none)	Required for custom. -70 to -5.
truePeakDbtp	number	-1	Ceiling, -9 to 0 dBTP (4× oversampled).
loudnessRange	number	11	Target LRA, 1–50 LU.
outputFormat	string	mp3	mp3, m4a, wav, flac, mp4 (mp4 needs a video input)
sampleRate	integer	same as input	8000–192000 Hz. MP3 is capped at 48000.
channels	string	keep	keep, mono, stereo
maxDurationSeconds	integer	14400	Longer inputs are rejected (free).

You must provide either url or files.

Output reference

Each file produces one dataset row. Here is a real 59-minute spoken-word test run (URLs shortened):

{
  "sourceUrl": "https://example.com/episodes/ep-142.mp3",
  "preset": "podcast",
  "outputFormat": "mp3",
  "status": "ok",
  "outputKey": "a1b2c3d4e5f6.mp3",
  "outputUrl": "https://api.apify.com/v2/key-value-stores/<storeId>/records/a1b2c3d4e5f6.mp3?signature=<sig>",
  "mime": "audio/mpeg",
  "target": { "integratedLufs": -16, "truePeakDbtp": -1, "loudnessRangeLu": 11 },
  "beforeLufs": -24.41,
  "afterLufs": -16.32,
  "afterTruePeakDbtp": -1.32,
  "gainAppliedDb": 8.09,
  "normalizationType": "dynamic",
  "renderPasses": 2,
  "truePeakCompliant": true,
  "input": {
    "format": "mp3",
    "duration": 3553.2,
    "sizeBytes": 28426087,
    "video": null,
    "audio": { "codec": "mp3", "sampleRate": 22050, "channels": 1 }
  },
  "billedMinutes": 60
}

Rows also include outputBytes, outputDurationSeconds, and full before and after objects (integratedLufs, loudnessRangeLu, truePeakDbtp, thresholdLufs). A failed file returns only sourceUrl, preset, outputFormat, status: "error" and error.

normalizationType is linear when plain gain was enough. It is dynamic when the limiter or range control had to work, which is usually the case for peaky audio or a large gain change.

Use it from AI agents and automations

MCP (Claude, Cursor, other MCP clients). Add https://mcp.apify.com?tools=ent-e/podcast-loudness-normalizer as a remote MCP server and sign in with OAuth. Your agent can then call the Actor with a URL and a preset. Long files take a while (roughly 8–9 minutes of processing per hour of audio), so the agent should check the run and read the dataset when it finishes.

n8n. Use the Apify node (@apify/n8n-nodes-apify):

• Run an Actor and get dataset for short clips.
• For full episodes, use Run an Actor, then Get Dataset Items.

Make. Use the Apify app:

• The Run an Actor module with Run synchronously waits up to 120 s, which suits promos and short clips.
• For episodes, run asynchronously. Use the Watch Actor Runs trigger, then Get Dataset Items.

Zapier. Use the Apify app:

• Run Actor and Fetch Results for runs under 5 minutes.
• Otherwise, use Run Actor, then the Finished Actor Run trigger and Fetch Dataset Items.

API. Use async for episodes. Start a run, then fetch the dataset when it succeeds:

curl -X POST "https://api.apify.com/v2/acts/ent-e~podcast-loudness-normalizer/runs?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com/episodes/ep-142.wav","preset":"podcast"}'
# then: GET https://api.apify.com/v2/datasets/<defaultDatasetId>/items?token=$APIFY_TOKEN

For short clips you can use /run-sync-get-dataset-items, which waits up to 5 minutes.

FAQ

Do I need a DAW, plugins or ffmpeg? No. Send a URL and get a normalized file and a loudness report back. Nothing to install.

Which preset should I use for Spotify, Apple Podcasts and YouTube?

• apple or podcast (-16 LUFS) for podcast feeds and Apple Podcasts.
• spotify or youtube (-14 LUFS) for music-style streaming targets.
• ebu-r128 (-23) or atsc-a85 (-24) for broadcast.

Why is my file at -14.6 LUFS and not -14? The true-peak ceiling always wins. If reaching -14 LUFS would push peaks above truePeakDbtp, the Actor stops short so the file doesn't clip on playback. The row shows the measured afterLufs and truePeakCompliant. If you want the full -14, allow more headroom (for example "truePeakDbtp": -0.5) or reduce the peaks in your mix.

Can I normalize a video podcast? Yes. Use "outputFormat": "mp4". The video stream is copied untouched, and only the audio is normalized and re-encoded to AAC.

Is the loudness measured on the final file? Yes. The after values come from measuring the encoded output, not from the filter's estimate. If the first render misses the ceiling or the target, it re-renders, and renderPasses tells you how many passes it took.

What isn't supported?

• Only the first audio stream is processed.
• Files must be direct download URLs (not streaming pages), up to 2 GB and 4 hours by default.
• It doesn't remove noise or EQ. It only sets loudness and peaks.

Am I charged if a file fails? No. Error rows are free. You pay only for files that were normalized and delivered.

ENT-E media toolkit

Every tool takes a public file URL, needs nothing installed, charges per file, doesn't charge for failures, and works from MCP, n8n, Make, Zapier and the API.

• Media Converter for AI Agents (FFmpeg): audio/video to MP3, Whisper-ready 16 kHz WAV, MP4, WebM, GIF or thumbnails, with trim and batch.
• Image Converter & Compressor: convert to WebP or AVIF, compress to a target file size, resize, strip EXIF/GPS.
• Podcast Loudness Normalizer: normalize to -16, -14, -23 or -24 LUFS with a true-peak ceiling, plus a before/after report.
• Video to Captions: SRT, VTT, TXT and word-level JSON from any video or audio URL, with Whisper.
• Video to Shorts: vertical 9:16 clips with burned-in animated captions for TikTok, Reels and Shorts.