If your Marathi videos will run as a YouTube playlist or feed a 24/7 channel, check each file’s loudness and true peak, then listen to the processed files and transitions. Normalisation can help manage level and peak differences, but it cannot repair distortion already present in a recording or guarantee that the sequence will sound even.
Marathi does not call for a special technical encoding workflow. The right settings depend on whether you are uploading videos to a playlist or sending a live encoder feed; those are separate delivery paths, and YouTube publishes different audio recommendations for each.
Upload playlist or live stream?
First decide what you are preparing. An uploaded playlist is a sequence of separate videos stored on YouTube. A live stream is a continuous feed sent to YouTube by an encoder. You may prepare video files before either route, but the platform settings and final checks are not interchangeable.
For ordinary uploads, YouTube’s recommended upload encoding settings specify an MP4 container, AAC-LC or Opus audio, and a 48 kHz sample rate. For stereo audio, the page lists a recommended bitrate of 384 kbps. These are upload recommendations, not a loudness target, and you should consult the current page for the full video settings.
Live encoder settings are on a separate YouTube Help page. It lists AAC or MP3 audio, 44.1 kHz for stereo and 128 kbps for stereo audio. It also recommends constant bitrate encoding and a two-second keyframe interval. Do not change a finished upload to the live values simply because you intend to play it during a stream: a file’s audio encoding and the encoder’s outgoing audio settings serve different stages.
| Delivery path | Audio guidance from YouTube | What to check |
|---|---|---|
| Uploaded video | AAC-LC or Opus; 48 kHz; recommended stereo bitrate 384 kbps | The rendered file, its processing, and playback on YouTube |
| Live encoder feed | AAC or MP3; 44.1 kHz stereo; 128 kbps stereo | Encoder output, stream preview, stream health, and archive |
The figures in the table are YouTube recommendations as listed on its upload and live encoder pages, checked for this article in October 2026. They are not interchangeable targets, and neither row sets a required LUFS value. If you are deciding how a prepared playlist becomes a live broadcast, the practical distinction is covered in this guide to making a live stream from a video playlist.
This article focuses on audio preparation. Keep your existing video workflow in view, but do not assume that choosing a compatible file profile will solve clipping. That requires checking the programme audio itself.
Inspect source clips for clipping
Before you alter levels, inspect each source. Listen to the loudest passages on headphones or speakers you know well. Pay attention to sharp consonants in speech, forceful vocal notes, percussion, bells, cymbals and dense musical sections. Look for a waveform with flattened tops or a meter that repeatedly reaches full scale, but do not rely on a visual display alone.
A peak meter indicates the highest digital sample level it sees. A signal that has reached full scale may already have been clipped in the original recording. Distortion can also be present in a source even when the file does not show an obvious flat-topped waveform. Listen for harshness, crackling, or a strained edge that follows the loudest sounds. If it is there before processing, turning the file down will make the distortion quieter, not remove it.
Make a simple source log. Record the filename, the passage that sounds loudest, any audible distortion, and the meter reading if your editor provides one. This prevents you from treating a bad source as a level-matching problem. If a recording is damaged, look for a cleaner version first. Repair tools may help in some cases, but any repair needs careful auditioning; aggressive processing can create a different sort of artefact.
Also note what kind of programme each video contains. A devotional song, a spoken introduction and a continuous instrumental piece can have similar peaks but feel very different in loudness. A single high peak does not tell you whether the whole item will sound much louder than the next one. That is why source inspection and loudness measurement belong together.
Measure loudness and true peak
Measure the finished programme audio of each clip, rather than adjusting by waveform height or peak meter alone. Integrated loudness estimates the overall level across a programme. Comparing it across items helps you spot a bhajan that may sound much louder than a spoken introduction, even if their sample peaks look similar.
True peak is a separate measure. It estimates peaks that can occur between digital samples during reconstruction or encoding, so it can reveal potential overload that a sample-peak reading misses. Integrated loudness helps with consistency from item to item; true peak helps you manage headroom. Neither measurement tells you whether a performance sounds natural, and neither replaces listening.
The FFmpeg filter documentation describes loudnorm as an EBU R 128 loudness normalisation filter. Its controls include integrated loudness (I), loudness range (LRA) and maximum true peak (TP). It supports single-pass processing and a two-pass approach for files. In dynamic mode, the filter upsamples internally to 192 kHz for true-peak analysis. That is a documented analysis behaviour, not a recommended export sample rate.
YouTube’s upload settings page does not specify a universal LUFS target. Do not treat a frequently repeated platform number as an official YouTube requirement unless you can find it in current primary guidance. Your appropriate programme target depends on whether the material is speech, music or a mix, as well as your intended dynamics and listening context. Use a consistent decision across the playlist, then check the results by ear.
If you need a published reference point, EBU R 128 Version 5.0 gives an average programme loudness recommendation of -23 LUFS for its broadcast context. The EBU document is not a YouTube upload specification. Use that figure only when you actually want to work to that broadcast recommendation; do not present it as a platform rule or a universal target for devotional and playlist content.
Normalise with a conservative ceiling
Once you have measured the clips, decide which ones need adjustment. If their integrated loudness is already close and they sound consistent in sequence, avoid processing for its own sake. If one item is clearly louder, reduce it; if another is much quieter, consider raising it while preserving headroom and the natural dynamics of the performance.
Set a maximum true-peak ceiling below full scale rather than pushing peaks right to the limit. A conservative ceiling leaves some margin for the rendered audio and subsequent encoding. The exact ceiling and loudness target are editorial choices, not numbers supplied by YouTube in the upload guidance. Choose settings appropriate to the material and verify the output with the same measurement method.
With a tool such as FFmpeg loudnorm, a two-pass process can use measured input values to guide linear normalisation when the filter’s constraints are met; otherwise it may use dynamic processing. A single pass is available too, but whichever mode you choose, inspect what it did. Dynamic processing can alter the relationship between quieter and louder parts. Linear gain is simpler when it can meet the requested target and ceiling, but it cannot satisfy an impossible combination of loudness and peak constraints without changing the signal.
Do not use a limiter as a way to make every item as loud as possible. Heavy limiting can flatten musical dynamics, make percussion tiring, and bring up background noise. The aim is dependable playback across the sequence, not maximum apparent volume. For a soft devotional recording, preserving a natural rise and fall may matter more than matching a loud, compressed intro exactly.
After processing, remeasure the rendered file and listen to its loudest passages again. Normalisation may lower a clipped file’s level, but it does not reconstruct the missing waveform. It also cannot guarantee freedom from every audible issue: clicks, bad edits, noise, codec artefacts, and abrupt transitions need their own checks.
Export using upload audio guidance
For each file intended as a YouTube upload, use YouTube’s upload recommendations as a practical starting point: MP4, AAC-LC or Opus audio, and 48 kHz. The upload page lists 384 kbps as its recommended stereo audio bitrate. Keep the frame rate aligned with the source recording and consult the full current encoding page for video-specific details rather than guessing from the audio settings.
These format recommendations are about compatibility and delivery. They do not decide how loud your programme should be. A file can use the recommended codec and sample rate while still having clipping, inconsistent levels, or a rough edit. Conversely, changing the sample rate to a live-stream value is not a remedy for audio peaks.
Export a short representative sample first if you are unsure about your editor’s settings. Check that the audio has not been muted, shifted out of sync, or rendered with an unexpected channel configuration. Then open the exported file in a player and listen, rather than assuming that the export dialog’s settings guarantee the result. When the playlist is a long-running part of a broader 24/7 channel workflow, keeping a clean set of upload-ready files also makes later review easier.
If instead you are encoding a live feed, configure that separately to YouTube’s live recommendations. Its live settings list AAC or MP3, 44.1 kHz stereo and 128 kbps stereo audio, alongside video encoder guidance. Keep the distinction clear in your notes: upload-file settings belong to each prepared video, while live settings describe the outgoing encoder feed. A playlist file’s sample rate does not, by itself, tell you how to set the encoder.
Listen across file transitions
Listen to the rendered versions, not only the original sources. For each item, check the opening, the loudest passage and the ending. This catches a fade that was cut off, an edit click, a sudden level jump, or a normalised section that now sounds unnaturally dense. Include quiet speech and singing as well as the loudest music; a peak-focused check can miss a quiet passage that is hard to understand.
Then play the playlist in its intended order and listen across every transition, or at least review the full sequence if it is short enough to do so. A clip may be technically unclipped and still sound startling after a much quieter one. Conversely, a transition that looks uneven on a loudness meter can sound acceptable because the content changes naturally. The listening decision is part of the work, not an optional polish step.
Check the end of one item and the start of the next without a pause. Some playback setups introduce a gap, while other workflows cut directly from one video to another. Listen for an unintended silence, overlapping sound, abrupt room-noise change, mismatched ambience or a sharp tonal shift. For a playlist of songs, also consider whether the sequence is meant to flow continuously or whether a clear break is appropriate.
If a transition feels wrong, first identify the cause. It may be an inconsistent level, a loud intro, an abrupt edit or a difference in recording quality. Adjust that cause rather than applying more compression to the whole playlist. Save a final sequence note with the version of each file you reviewed, so you do not accidentally upload an earlier export.
A file-based workflow has the advantage that you can review each item before it is public. It also asks you to allow time for processing and checks. If you are deciding between an uploaded playlist and a continuous broadcast, the guide to running a 24/7 devotional YouTube channel covers the broader format decision without changing the audio checks here.
Test the final YouTube result
An export that sounds fine locally still needs a platform check. For an upload playlist, upload a sample or the finished files privately or as unlisted videos, then wait for YouTube to process them. Play them back on YouTube and listen to the same loud and quiet passages and transitions you checked in the rendered files. YouTube’s upload and publish guidance distinguishes uploading from publishing, so leave time to review before making the sequence public.
Check the playback you will actually rely on: the device, speakers or headphones you use to monitor your channel, and the playlist order viewers will hear. Compare the online playback with your local render. If a problem appears only after upload, confirm that you are watching the processed version and that you have selected the correct file before changing the master. Keep an unmodified copy of the export so you can compare revisions.
For a live stream, test the encoder feed before starting the real broadcast. YouTube advises testing in advance, including audio and movement similar to the planned stream. Check the preview, listen for level or sync problems, and monitor stream health during the test. If you keep a stream archive, verify that it plays back correctly too.
A prepared playlist can be sent through an always-on live workflow, but that changes the final test, not the underlying need to prepare clean files. If maintaining a broadcast depends on your own computer staying on, that is a separate operational concern: StreamNeo can remove that particular burden by letting you hand off an uploaded file and channel key while your computer is off. It does not replace audio review, make a damaged recording clean, or change YouTube’s requirements.
Before committing, compare the operating options on the pricing page. When the file and channel are ready, start free — 24-hour trial, no card.
FAQ
Does Marathi need different audio settings for YouTube?
No special Marathi-specific encoding workflow is identified in the cited YouTube upload or live-stream guidance. Use the settings for your delivery path and check the audio in each file; language does not remove the need to inspect its source quality, loudness and peaks.
What LUFS target should I use for a YouTube playlist?
The cited YouTube upload settings do not specify a universal LUFS target. Choose a consistent programme level and true-peak ceiling that suit your material, then measure and audition the rendered files. EBU R 128’s -23 LUFS figure is a broadcast recommendation, not a YouTube requirement.
Will normalisation fix a clipped recording?
No. Normalisation changes level and can help manage peaks, but it cannot restore audio already distorted by clipping in the source. Find a cleaner recording if possible; if you attempt repair, listen carefully to the result.
Should I use upload settings for the live encoder?
No. YouTube lists different audio recommendations for uploaded files and live encoder output. Prepare the files for upload separately, then configure and test the encoder feed against the current live-stream guidance.