ToneParcel — Audio Mastering & Loudness API

Batch basic mastering with six presets, optional sound controls, loudness normalization, and measured output. Analyze or normalize files without applying the mastering effects when desired.

Input

{
  "urls": ["https://example.com/your-audio.wav"],
  "options": {
    "mode": "master",
    "preset": "warm-tape",
    "targetLufs": -14,
    "truePeakCeiling": -1,
    "format": "wav",
    "effects": { "warmth": 45, "stereoWidth": 105 }
  }
}

Replace the example URL with your public WAV/MP3 file. Maximum 10 URLs per run, 0.4–300 decoded seconds, 60 MiB per file, mono/stereo, and 44.1/48 kHz input. WAV must contain PCM audio. Output supports 44.1/48 kHz WAV (16/24-bit) or MP3 (128/192/256/320 kbps). Files are processed sequentially.

Option	Default	Meaning
mode	master	master, normalize or analyze
preset	transparent	Mastering preset; only accepted in master mode
targetLufs	Preset target, otherwise -14	-30 to -9 LUFS
truePeakCeiling	Preset ceiling capped at -1	-6 to -0.3 dBTP
format	wav	wav or mp3
sampleRate	48000	44100 or 48000
bitDepth	24	16 or 24; WAV only
mp3Bitrate	192	128, 192, 256, 320; MP3 only
effects	Preset values	Optional overrides in master mode

Presets: Transparent, Warm Tape, Crystal Air, Punch Glue, Wide Cinema, Loud & Clear. A preset sets a processing configuration; individual effects then override it. There is no genre classifier, model inference, stem separation or reference-track matching.

Available effect overrides: warmth, clarity, air, lowEndClean, glueCompression, autoLevelStrength (0–100), stereoWidth (0–150; 100 keeps the original width), and monoBassAnchor (boolean). Stereo controls have no effect on mono inputs. The existing preset's depth setting remains part of its sound. Transparent applies mild processing; use mode: "normalize" to avoid mastering coloration.

Processing and output

The mastering DSP runs in a Node worker using its export chain. FFmpeg decodes inputs, applies final two-pass loudness normalization and encodes the result. Metered input/output LUFS and true peak come from FFmpeg. MP3 is measured after encoding; if encoding creates excess peaks, the output is attenuated and measured again. The returned target_met indicates whether actual LUFS is within 0.5 LU of the target. Peak protection can take precedence and leave output quieter than requested.

The Dataset returns index, status, duration, channels, actual settings, engine version, input/output measurements, billable_minutes, billing_event, and output_url/output_key for processed audio. Analysis returns measurements without an audio output. Silent/below-gate input returns lufs: null and measurement_status: "unmeasurable" in analysis; normalization/mastering rejects it. Sub-400ms inputs are rejected rather than assigned an invented LUFS value.

Per-file errors are free of result charges, and other files continue. If all files fail, the run fails. SUMMARY records outcomes and acknowledged billing events. Read the measured output and warnings rather than assuming every target is reachable. Synthetic regression tests do not establish subjective mastering quality or standards certification.

Downloads permit public HTTP(S) URLs only, reject private/reserved addresses and pin DNS results. Files are downloaded into a random per-item temporary directory and removed afterward. The DSP worker is terminated after 300 seconds; each FFmpeg stage has a 120-second deadline. Output metadata is stripped. No input audio is sent to an external AI service.

Output URLs refer to shareable records in the run's Apify key-value store; anyone holding a URL may be able to download it. Retention follows Apify store settings, without guaranteed fixed-hour deletion or permanent hosting. Save results to your own storage. Input JSON and result rows remain subject to platform retention.

Launch pricing (USD, platform usage included)

Mode	Each started minute, per successful file
Analyze	$0.010
Normalize	$0.020
Master, any preset	$0.040

Each run also costs $0.005, including runs with failed files. Duration is rounded up separately for each file: 60 seconds = 1 minute, 60.1 seconds = 2. A 3-minute master is $0.125 total, and a 5-minute master is $0.205. No subscription or preset surcharge.

Each file generates one event for its mode and 1–5-minute duration tier. This avoids charging only some minutes of an indivisible output. Budget checks occur before download and after decoded duration is known; insufficient budget stops the batch without delivering or charging that file. Preparation can incur compute before its exact duration is known.

The fixed 2 GB run uses two automatic start events at $0.0025 each, totaling $0.005. There is no separate Dataset-row charge.

Call from your application

Start an asynchronous run using your Apify token and an explicit spending limit:

curl -X POST "https://api.apify.com/v2/acts/epubpreflight~toneparcel-api/runs?maxTotalChargeUsd=1" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  --data-binary @input.json

Save data.id from the response. Poll GET /v2/actor-runs/{runId} with the same authorization header until the run finishes. Read GET /v2/datasets/{defaultDatasetId}/items to obtain result rows and download each successful output_url. Analysis-only runs return measurements without an audio file.

Use only source files you have permission to process. For help, open an issue on this Actor and include the run ID, without posting confidential source URLs.