Master a long audiobook in one file, up to 2 hours
An audiobook, a podcast or a lecture rarely fits into 10 minutes. Until now such a recording had to be cut into parts and mastered one by one, and then you had to make sure the parts matched in loudness. Now Magic Master takes a recording up to 2 hours in one file.
How it works
- Upload the file to mastering. If it is longer than 10 minutes, the app shows the length, the processing time and the price: for 47 minutes, about 8 minutes and 4 tokens. The price and the time are shown before you start.
- Pick the type: Voice, Podcast, Audiobook or Music (DJ set, live show).
- Pick the output format and press "Process". You can close the tab: you will get an email, and the result stays in the app under "My long tracks".
Long recordings need an account, so the result and the tokens are tied to you. You can cancel at any time and the tokens come back.
Why the loudness does not drift
The server processes a long recording in one-minute parts, so it never has to hold the whole file in memory. To keep the parts sounding the same:
- The whole file is analysed first. Loudness (LUFS) and the noise profile are measured once on the entire recording, not per part.
- One set of settings for every part. The EQ, the phrase leveller and the compressor are the same from the first minute to the last.
- A run-up before each part. Processing of each part starts a little before the part itself, so the filters and the compressor settle exactly as they would in one continuous pass.
- Joins in pauses, with a crossfade. A join goes to the quietest moment nearby, is refined to a zero crossing, and a 4-second overlap is blended with a smooth crossfade.
- One final loudness. The final gain and the limiter run over the joined recording as a whole: integrated loudness lands on the target and the peak stays below the limit.
In our tests on speech, noise and a sine wave, the joined result matched one-pass processing of the file to below −80 dB, which is inaudible.
Modes and loudness
| Mode | Loudness | Peak | What it does |
|---|---|---|---|
| Voice | −16 LUFS | −1 dBTP | noise reduction from the noise profile of the whole recording, even phrase level |
| Podcast | −16 LUFS | −1 dBTP | the same, gentler |
| Audiobook | −19 LUFS | −3 dBTP | ACX guide: RMS −18 to −23 dB, peak no higher than −3 dB |
| Music | −14 LUFS | −1 dBTP | DJ set, live show, an album in one file, no noise reduction |
More on audiobook requirements: Audiobook mastering online.
Price
- Up to 10 minutes: 1 token, like regular mastering.
- Every started 15 minutes after that: 1 more token.
- For example, 47 minutes is 4 tokens and 2 hours is 9 tokens.
The free daily mastering does not cover long recordings: they take far longer to process than a song. If processing fails or you cancel it, the tokens come back.
Formats
Input: WAV, FLAC, MP3, M4A, AAC, OGG, Opus, AIFF, WMA and WebM. For long recordings FLAC or MP3 is handier: two hours of WAV is over a gigabyte. Output: WAV 24-bit, FLAC or MP3.
Questions and answers
How long can the file be?
Up to 2 hours in one file. If the recording is longer, split it into parts, for example by chapter.
How much does it cost?
Up to 10 minutes: 1 token, like regular mastering. Every started 15 minutes after that: 1 more token. The app shows the exact price before you start, right after the upload.
Will the loudness jump where the parts are joined?
No. Loudness and tone are measured once on the whole file, every part uses the same settings, and the joins are smoothed with a crossfade. In our tests the joined result matched processing the file in one pass to below −80 dB.
Can I close the tab while it is processing?
Yes. Processing runs on our server. When it is done you get an email, and a Telegram message if Telegram is linked. The result stays in the app for 72 hours.
What if processing fails?
The tokens go back to your balance automatically. The same happens if you cancel.
Upload a track — the same mastering engine, prices in euros.
Open Echoteka →AI mastering onlineLUFS analyzer