How to make rain sound videos for YouTube

1. What people search for in a rain video

Google's search suggestions on November 4, 2026 show two groups of searches. Creators ask "how to make rain sound videos on youtube", "are rain sound videos monetized on youtube" and "how much do rain sound videos make on youtube". Viewers ask for a kind of rain: "rain sounds 10 hours no ads", "rain sounds black screen", "rain and thunder sounds for sleeping", "heavy rain sounds live", "gentle rain sounds live", "rain sounds live 24 7", "thunderstorm sounds for sleep black screen". Each viewer search is a separate video.

The general choices for any sleep sound channel, a long upload or a live stream, the repetitive content rule and what noise channels are documented to earn, are in our guide to white noise and sleep sounds channels. This guide covers what is different about rain.

2. Recording rain yourself

Rain is one of the few sleep sounds anyone can record at home, and a recording of your own is the thing YouTube's rules reward most (section 7). What rain sounds like depends almost entirely on what it lands on, so the surface is the first choice:

• A window: soft taps and runs of water; record from inside with the window shut, which keeps the recorder dry and cuts some street noise.
• A tin or metal roof, a car roof, a shed: loud, bright drumming. Viewers search for them by name: "rain on tin roof", "rain on tent", "rain on window sounds".
• A tent or a tarp: close, papery patter. Record from inside the tent.
• Leaves, a forest, a garden: wide and soft, with drips. Needs the recorder outside, so it needs cover.
• A gutter or downpipe: a steady stream under the rain, useful as a second layer.

Practical points, from the way field recordings usually go wrong:

• Keep the recorder dry, and keep its cover quiet. A raindrop hitting the microphone itself, or a plastic bag over it, sounds like a tap on the mic, not rain. Put it under an overhang, an umbrella fixed above it or an open window, so the drops land near it, not on it.
• Use a windshield outside. Gusts on a bare microphone make low thumps that a later filter cannot fully remove.
• Set the level on the thunder, not the rain. Rain is steady and quiet; a close thunder clap is far louder. If the level is set for the rain, the thunder runs into the top of the recorder's range and the waveform is cut flat. In our test, raising a generated 8-second thunder sound 10 dB past its safe level left 39,740 of its samples stuck at full scale; a limiter applied later cannot rebuild those tops. Leave room, and turn the rain up afterwards.
• Listen for people. Traffic, a neighbour's television, voices and dogs get louder in your headphones than you remember them. A late night or early morning recording, with the phone in flight mode, avoids most of it. Cut any stretch where words can be understood: YouTube's privacy rules apply to voices as well as faces, as our coffee shop ambience guide explains.
• Record far longer than you think. An hour of real rain gives you a long loop with thunder that never repeats in the same place; a few minutes gives you a loop listeners notice.

Write down where and when you recorded and on what. That note is the start of a description that tells viewers why this rain is not the same file as the next channel's.

3. "Copyright free" rain sounds: what the licences say

"Rain sounds copyright free", "rain sound free for commercial use" and "rain sounds copyright free download" are common searches. "Copyright free" is not a licence. What decides whether you can put a rain file in a monetized video is the licence on that particular file, and libraries mix licences.

On Freesound, the licence filter on a search for "rain" offered Creative Commons 0, Attribution, Attribution NonCommercial and the older Sampling+ on November 4, 2026, so check the licence line on every sound page, and keep a copy of it with the file. Freesound requires an account to download.

Content ID and rain. Plain rain is hard to claim through Content ID: YouTube lists "Sound effects, soundbeds, or production loops" as ineligible to be used as a reference. That protects you less than it sounds. Freesound's FAQ describes how a free sound still gets matched: if an artist uses an unchanged sound in a song that is registered with Content ID, "if another person uses the same sound effect in a YouTube video, the ID scan will discover a match." Its advice is to dispute with "evidence that the Freesound upload predates the copyrighted work", and it notes that "If the sound was modified with plugins, stretched or slowed, etc., then the system is less likely to find a match." Layering your own recording with library sounds helps on both counts.

YouTube adds: "Only music and sound effects from the Audio Library are known to YouTube to be copyright-safe. YouTube is not responsible for issues that arise from “royalty-free” music and sound effects from YouTube channels or other music libraries." A video on another channel that calls its rain free to use is not a licence.

4. Add thunder at random times

A thunder file looped every few minutes is the easiest rain video to spot: the storm arrives on a timetable. Placing each clap at a random time, at a random level, fixes that. This script mixes one hour of rain (rain-1h.wav) with 12 copies of one thunder recording (thunder.wav):

n=12
times=$(shuf -i 20-3580 -n $n | sort -n)
f="[1:a]asplit=$n$(for i in $(seq $n); do printf '[s%d]' $i; done);"
mix="[0:a]"; i=0
for t in $times; do
  i=$((i+1)); v=$(shuf -i 0-9 -n 1)
  f="$f[s$i]volume=-${v}dB,adelay=${t}s:all=1[t$i];"
  mix="$mix[t$i]"
done
echo "thunder at:" $times
ffmpeg -i rain-1h.wav -i thunder.wav -filter_complex \
  "$f${mix}amix=inputs=$((n+1)):duration=first:normalize=0" -c:a pcm_s16le storm-1h.wav

• shuf -i 20-3580 -n 12 picks 12 different whole seconds between 20 and 3,580, so no clap is cut off at the start or the end of the hour.
• asplit makes 12 copies of the thunder; volume=-${v}dB makes each one 0 to 9 dB quieter than the original, so the storm comes nearer and goes further away.
• adelay=${t}s:all=1 moves each copy to its time on every channel. amix with duration=first keeps the length of the rain, and normalize=0 keeps the rain at its own level instead of dividing everything by 13.
• Better still: use several different thunder recordings and pick between them, since 12 copies of one clap can be recognised.

What we tested: on November 4, 2026, with ffmpeg 7.0.2, we ran the script on an hour of generated rain (pink noise, filtered) and an 8-second generated thunder rumble. The result was exactly 3,600 seconds long. Measuring the loudness every 100 ms found 12 thunder events, each starting at the second the script printed. One run placed two claps 10 seconds apart; random times can bunch up, so listen to the result and run the script again if you do not like the spread.

5. Thunder that does not wake the listener

Most people play a thunderstorm video to fall asleep. A real thunder clap is many times louder than the rain under it, and played at the volume that makes the rain comfortable, it can wake the person it was meant to help. Thunder should be heard, not felt. We tested four ways to tame it on a one-minute mix of our test rain with three thunder claps.

The limiter only shaves the top of each wave: the peak dropped by almost 4 dB, but the clap was nearly as loud as before. The two normalisers smooth loudness over time and are too slow for a sudden clap. The compressor reacts within milliseconds and leaves anything below its threshold alone, which is exactly what thunder over rain needs:

ffmpeg -i storm-1h.wav -af \
  "acompressor=threshold=-20dB:ratio=4:attack=5:release=1000,alimiter=limit=0.5:level=0" \
  -c:a pcm_s16le storm-tamed.wav

• threshold=-20dB: set it at the loudest level of your rain alone. Find that with ffmpeg -i rain-1h.wav -af volumedetect -f null - and read max_volume; ours was -20.6 dB. Rain under the threshold passes through untouched (our rain-only test came out unchanged).
• ratio=4: above the threshold, every 4 dB of thunder becomes 1 dB. A higher ratio flattens the storm: at threshold -24 dB and ratio 8 our thunder was only 1.5 LU above the rain, barely a storm.
• attack=5:release=1000: react in 5 ms, let go over a second, so the rumble's tail fades naturally.
• alimiter=limit=0.5:level=0 catches any peak the compressor let through and stops the limiter from turning the whole mix back up. The ffmpeg manual says it "prevents an input signal from rising over a desired threshold".

On the full hour from section 4, the same command brought the loudest thunder from 18.8 LU to 6.5 LU above the rain, and the highest peak from -2.3 to -14.0 dBTP. Our thunder was generated; a real recording has a sharper crack, so listen on headphones at bedtime volume and adjust the threshold and ratio by ear.

6. Build the 10-hour file

For "rain sounds 10 hours", loop your hour of rain to the length you want. The method is the one in our ambient video guide: keep the sound uncompressed until the last step, and encode it once. First crossfade the end of the hour into its start, so the loop point cannot be heard:

ffmpeg -ss 3 -i storm-tamed.wav -t 3 -i storm-tamed.wav -filter_complex \
  "[0:a][1:a]acrossfade=d=3" -c:a pcm_s16le storm-loop.wav

The result is 3 seconds shorter than the hour (3,597 seconds in our test). If a thunder clap falls in the first or last few seconds, it will be cut by the fade; the script in section 4 keeps them away from both ends. Then make 10 seconds of black picture, and repeat both to 10 hours:

ffmpeg -f lavfi -i "color=c=black:s=1920x1080:r=30" -t 10 \
  -c:v libx264 -tune stillimage -preset veryfast -pix_fmt yuv420p black-10s.mp4

ffmpeg -stream_loop -1 -i black-10s.mp4 -stream_loop -1 -i storm-loop.wav \
  -map 0:v -map 1:a -t 36000 -c:v copy -c:a aac -b:a 192k -movflags +faststart rain-10h.mp4

The picture is copied, not re-encoded, so the job is fast; only the sound is encoded, once. Replace black-10s.mp4 with your own rain footage loop for a picture (section 7).

What we tested: the same command with -t 7500 instead of 10 hours, so it crossed the loop point twice. The file had no silent gap at either join and its index was at the start. For comparison, encoding the hour to AAC first and repeating that with -c copy left about 13 ms of silence at each join. Our guide to seamless loop videos explains where that gap comes from.

A black picture costs almost nothing, so nearly all of the file is sound. Filmed rain is different: at YouTube's recommended 8 Mbps for 1080p, the picture alone is 3.6 GB an hour (our arithmetic), and our ambient guide has a table of sizes by resolution. All of these fit easily under YouTube's limit of "256 GB or 12 hours, whichever is less" for verified accounts.

7. Black screen or a dim picture?

"Rain sounds black screen" is one of the most suggested searches, and the reason is practical: a phone or TV left on all night lights up the room. A black screen solves that. It also gives the video nothing but its sound, which matters for monetization.

YouTube's monetization policies do not mention black or still screens. The nearest wording is in the examples of what is "not allowed to monetize": "Similar or repetitive content with low educational value, commentary, narratives, or minimal variation across videos" and "Image slideshows, templated storylines, or scrolling text with minimal or no narrative, commentary, or educational value". The policy says it looks at "the substance of each video", which for a black-screen channel is the sound alone. So the sound has to carry the difference between videos: a different place, a different storm, a different surface. Ten black-screen uploads of one file with new titles are what that wording describes. Our guide to inauthentic content and 24/7 channels goes through the policy line by line.

A middle way is a dim picture: your own rain footage, darkened so it shows the place without lighting the room. This filter lowers brightness and pulls the mid-tones down:

ffmpeg -i rain-clip.mp4 -vf "eq=brightness=-0.15:gamma=0.6" \
  -c:v libx264 -preset veryfast -pix_fmt yuv420p -an rain-clip-dim.mp4

On a bright 5-second test picture it lowered the average brightness from 126 to 53 on the 0 to 255 scale, about 42% of the original (our arithmetic). Make the loop of the dimmed clip as in our ambient guide, then use it in place of black-10s.mp4 in section 6. Many channels publish both: a picture version and a black-screen version of the same recording, which serves two searches without being the same video twice under two titles.

8. "No ads" in the title

"Rain sounds 10 hours no ads" and "rain sounds live no ads" are among the most suggested searches. A creator cannot guarantee it. YouTube's Terms of Service say: "You grant to YouTube the right to monetize your Content on the Service (and such monetization may include displaying ads on or within Content or charging users a fee for access). This Agreement does not entitle you to any payments." A title that promises no ads can be wrong the day YouTube places one, and viewers who are woken by an ad tend to say so in the comments. Our aquarium video guide covers the same promise for "no ads" aquarium videos.

What you can control, once you are in the Partner Program, is the ads in the middle. YouTube says: "On monetized videos that are 8 minutes or longer, you can turn on ads during the middle of the video (known as "mid-rolls")." For a sleep video, a break that lands at 2 a.m. is the one that wakes people. Leaving mid-rolls off is a choice you can describe honestly, for example "no mid-roll ads", where "no ads" is not.

9. A 24/7 rain stream

"Rain sounds live 24 7", "rain and thunder sounds live" and "rain sounds black screen live" show that viewers also look for rain in the Live tab. The trade-offs between a live stream and long uploads, including the 12-hour archive limit ("If your stream exceeds 12 hours, it may not be captured at all") and what that means for watch hours, are the same as for any sleep sound and are compared in our sleep sounds guide. Two points are particular to rain:

• A loop is heard more often live. A viewer who keeps a rain stream on every night hears the same thunder at the same moment every hour. Build the loop from several hours of different recordings, or several runs of the random thunder script, rather than one hour repeated.
• Thunder and no thunder are two streams. People who want "rain sounds no thunder" leave at the first clap. Run them separately, or keep the stream to one and say which in the title.

Say in the description that the stream is a recording played on a loop, not live weather. Rain streams sit with the other always-on formats in our list of 24/7 live stream ideas, and a rainy window often pairs with a fireplace video in the same channel.

10. Do rain sound videos make money?

"How much do rain sound videos make on youtube" is a common search, and there is no reliable published answer for YouTube. Figures for named rain channels come from estimator sites that multiply public views by a guessed rate; we do not repeat them. The documented evidence for noise and sleep audio, mostly from Spotify, and why Spotify has since cut what noise earns, is in our sleep sounds guide, and YouTube revenue shares and the Partner Program thresholds are in do lofi channels make money.

Two rain-specific facts shape the money side. Rain sound videos can be monetized like any other video, if the channel passes the repetitive and reused content review in section 7. And because YouTube does not accept "Sound effects, soundbeds, or production loops" as Content ID references, you have no automatic way to collect on copies of your rain; a recording with a clear place and story is easier to defend with a manual copyright request than an anonymous loop.

Before you upload

1. The rain is your own recording, or every file's licence page is saved and allows commercial use; CC BY sounds are credited in the description.
2. No BBC Sound Effects or "Noncommercial" sounds are in a monetized video without a paid licence.
3. No words, voices, music or television can be heard in the recording.
4. Thunder was checked at bedtime volume on headphones; nothing jumps out.
5. Thunder times are irregular, and the loop point is crossfaded.
6. The long file was built from WAV and encoded once; it is under 12 hours.
7. The title says "no thunder" or "with thunder", and does not promise "no ads".
8. Each upload is a different recording, not the same file with a new title or length.
9. A live stream says it is a recording on a loop.

FAQ