How to resize an image

1. Choose an image or drag one onto the box. Its pixel dimensions and file size appear in the table.
2. Enter a new Width or Height in pixels, or pick a Quick scale. Original size puts both back.
3. Turn off Keep proportions if you need a shape that doesn't match the original. The picture is stretched to fit.
4. Check the new pixel size and file size, then press Download PNG.

Shrinking large photos

When you scale an image down to less than half its size, the tool halves it in steps before the final resize instead of jumping straight to the target. One large jump can leave jagged edges on fine detail, and stepping down keeps the result smooth. Making an image bigger works too, but it can't add detail that isn't in the original. Because the output is PNG, a resized photo can have a larger file size than the JPEG you started with. If that matters, run it through the Image Compressor afterwards. An animated GIF is read as a single still frame.

How to compress an image

1. Choose an image.
2. Pick JPEG or WebP under File type.
3. Lower Quality until the new size looks right. It starts at 75. The preview shows the compressed result; press Show original to compare.
4. Press Download JPEG or Download WebP.

JPEG or WebP

WebP usually gives a smaller file than JPEG at a similar quality, and it keeps transparent areas. JPEG opens in almost any software, but it has no transparency, so transparent pixels are filled with white. If the result says the file got larger, the original was already compressed harder than your setting, so lower the quality or keep the file you have. Low settings show first on text and fine lines, so check the downloaded file up close before using it on artwork with lettering. If your browser can't create WebP files, the WebP choice is switched off and says so. An animated GIF is read as a single still frame.

How to convert an image

1. Choose an image or drag it onto the box.
2. Pick PNG, JPEG or WebP under File type.
3. If the image has transparent areas and you picked JPEG, choose the color that fills them.
4. Check the new file size and press Download.

Which file type to pick

PNG is lossless, which makes it the right choice for logos, screenshots and anything with transparency or sharp text, at the cost of bigger files. JPEG suits photos that need to open anywhere, as long as you don't need transparency. WebP keeps transparency and usually produces smaller files than either, which makes it a good default for images on the web. If your browser can't create WebP files, the WebP choice is switched off and says so. For BMP, ICO favicons, PDF, HEIC, TIFF, AVIF, camera RAW or converting a batch of files at once, use the File Converter. An animated GIF is read as a single still frame.

How to crop a photo to a ratio

1. Choose an image.
2. Pick an Image shape: Square (1:1) for cover art and profile photos, Portrait post (4:5) for feed posts, Story (9:16) for stories, Widescreen (16:9) for video thumbnails, or Photo (3:2) for a standard photo print. Each one shows the size it will produce.
3. Drag the frame in the preview, or use the Position slider, to choose which part of the picture to keep.
4. Check the cropped size and press Download PNG.

Getting an exact size

The crop keeps every pixel inside the frame, so the output is as large as your picture allows. A 4000 × 3000 photo cropped to 1:1 comes out at 3000 × 3000. If a platform asks for a specific size, such as 1280 × 720, crop to the shape here first and then set the exact dimensions in the Image Resizer. When the picture already has the chosen shape, the page says so and there's nothing to move. An animated GIF is read as a single still frame.

How to remove a background

1. Choose an image, drag one onto the page, or paste one with Ctrl+V (⌘V on a Mac). JPG, PNG and WebP files work.
2. Wait for the result. The first time, your browser downloads the background remover (about 98 MB); it keeps it, so later images start sooner.
3. Drag across the image to compare the original with the result. Switch the preview to white, black or your own colour to check the edges. The checkerboard only marks transparent areas and isn't part of the file.
4. If something was missed or cut away, pick Erase or Restore and paint over it. Undo and Redo step through your strokes.
5. Press Download PNG. To get the cut-out on a solid colour instead, pick that colour as the preview background and use Download with this background.

How it works

The page runs BiRefNet_lite, an open image segmentation model released under the MIT licence, through Transformers.js. Where your browser supports WebGPU it uses your graphics card; otherwise it runs on the processor with WebAssembly, which works in every current browser but can take 20 seconds or more per image. The model looks at a 512 × 512 copy of your picture. Its outline is then lined up with the edges of a sharper copy, up to 2048 pixels, with a guided filter, and scaled to the full size when you download. The PNG keeps your picture's own pixels; only transparency is added.

What gets downloaded, and what doesn't

Your image is never sent anywhere: decoding, the model, the touch-ups and the PNG all happen on your device. The first time you use the tool, the page downloads the model file from Hugging Face and the ONNX Runtime engine from jsDelivr, and your browser stores both, so later visits only download them again if the browser has cleared its storage.

Limits

Pictures up to 50 megapixels open, and the PNG keeps the original size up to 50 megapixels (16.7 on iPhone and iPad, where Safari can't build a larger image; the page tells you when it scales down). It works best with one clear subject. Fine hair against a busy background, glass, smoke, and subjects the same colour as their background are the hardest cases, and the brushes are there for those.

How to make a profile picture

1. Choose a photo. It starts centered.
2. Drag Zoom to tighten the framing, then drag the preview (or use the position sliders) to move the subject into place. Reset framing puts it back.
3. Pick Circle or Square.
4. Pick an Export size: 256, 512 or 1024 px, then press Download PNG.

Circle or square

Many apps already display avatars as circles, so a square upload is often enough and leaves the app room to frame it. Circle is useful where nothing will crop the image for you, like a website, a stream overlay or a video thumbnail, because the area outside the circle is saved as transparent pixels rather than a solid color; the preview shows those corners as a checkerboard. If the framed part of your photo is smaller than the size you pick, it gets scaled up and will look soft, and the page tells you when that happens. An animated GIF is read as a single still frame.

How to get hex codes from a photo

1. Choose an image.
2. Set the Number of colors you want.
3. Click a color under the image to copy its hex code, or press Copy all.
4. Press Download hex codes for a .txt file with one hex code per line.

How the colors are picked

The image is scaled down to a small sample and every pixel is sorted into one of 4,096 color groups. The groups with the most pixels win, and each swatch is the average of the real pixels in its group, so it's a color that actually appears in the image. That means the palette favors area: a large grey sky will outrank a small red logo. To get an accent color that only covers a small part of the picture, crop in on it first with the Image Cropper. Transparent pixels are ignored, and two close shades can both appear when a color falls on the edge between groups. An animated GIF is read as a single still frame.

How to edit a photo

1. Choose a photo.
2. Drag the sliders under Light and color first, then open Effects for black & white and blur. Brightness, Contrast and Saturation start at 100%, which leaves the photo unchanged. Each slider has its own Reset.
3. Press Show original to compare, then press it again to go back to your edit.
4. Press Download PNG. Reset adjustments puts every slider back; Start over also clears the photo.

Size and how the effects work

Photos wider than 1600 pixels are scaled down to 1600 pixels wide for editing and export, which suits social posts and the web but not large prints. The effects use the same filter engine as CSS, so the download matches the preview. Warmth works by blending in a sepia tone, so high values push the whole image toward brown. Blur is measured on the scaled image, which means the same setting looks stronger on a small photo than on a large one, and strong blur can leave the edges slightly see-through. An animated GIF is read as a single still frame.

How to convert an audio file

1. Choose an audio file. The Original line shows its format, sample rate, channels and length.
2. To keep everything as it is, press Download WAV.
3. To change the sample rate or switch between mono and stereo, open Change sample rate or channels first.
4. Check the file type, sample rate, channels, length and size listed above the button, then press Download WAV.

Which sample rate to pick

44.1 kHz is the CD standard and the usual rate for music releases. 48 kHz is standard for video, so match it when the audio is going into a video edit. 22.05 kHz halves the file size compared with 44.1 kHz but loses the high end, which is acceptable for voice notes and rough demos. Switching stereo to mono combines both channels into one and halves the size again. This page only writes WAV. To convert audio to MP3, FLAC, M4A, 24-bit WAV or AIFF, use the File Converter.

The file is decoded at its own sample rate, read from the file header, so Keep really keeps it. When the rate or channels change, your browser’s own resampler does the conversion. A WAV is uncompressed: 16-bit stereo at 44.1 kHz takes about 10 MB a minute, which is why the size is shown before you download.

How to remove silence from a recording

1. Choose an audio file. The tool shows how much silence it will remove from each end, and the Original line shows the file’s length and peak level.
2. If needed, set the Silence threshold, from -60 to -20 dB. Anything at the edges quieter than this gets cut.
3. Set Edge padding, up to 500 ms, so the audio doesn't start or stop abruptly. Reset adjustments puts both back to -45 dB and 80 ms.
4. Press Play trimmed audio or Play the last 3 seconds to check the cut, then press Download WAV.

Choosing a silence threshold

The default of -45 dB suits most clean recordings. If background hiss or room noise is being kept, raise the threshold toward -30 dB so the noise counts as silence. If the soft start of a note or a breath is getting clipped, lower it toward -60 dB or add more padding. The cut is made at the first and last samples louder than the threshold on any channel, so a single click or pop near the edge will stop the trim at that point.

The threshold is in dB below full scale (0 dB is the loudest a file can store). The waveform is only a picture; the play buttons and the WAV both use the exact samples between the two pink markers, at the file’s own sample rate.

How to normalize an audio file

1. Choose an audio file. The Original line shows its current peak and average level.
2. Under Match on, pick Peak or Average (RMS).
3. Set the Target level. It starts at -3 dB peak, a safe default that can't clip; Reset adjustments goes back to it.
4. Check the gain being applied, the new peak and average, and the clipping line, then press Download WAV.

Peak or average

Peak moves the loudest single sample to the target, so the result can't clip. It's exact, but two files with the same peak can still sound quite different in loudness. Average matches the RMS level instead, which lines up how loud clips feel more closely. Raising the average can push peaks past full scale, so the gain is capped to keep the loudest sample at -0.1 dB, and the page tells you when clipping protection has limited the gain. This isn't a LUFS measurement, so don't rely on it to hit a streaming platform's loudness target.

The peak is found by reading every sample on every channel, so a single short transient is never missed. The same gain is applied to every sample, and the WAV keeps the file's own sample rate and channels.

1. Choose a MIDI file.
2. Under Soundtrack, choose an audio bounce of the same song if you want sound, and pick a Video size: 720p, 1080p, Vertical or Square.
3. Under Tracks, untick any you don't want and pick a colour for each.
4. Adjust Fall time, and open Note style, Glow or Background and keyboard for finer changes.
5. Press Export video, let it play to the end, then press Stop and save.

Tempo changes inside the file are followed, so slowdowns and section changes stay in time. With a soundtrack attached, the roll follows the audio's clock, and the audio starts at the very beginning of the MIDI file, so export your bounce from the start of the project too. Recording runs in real time at 30 frames per second, so a three-minute piece takes three minutes, and the video saves as WebM with no watermark. Files that use SMPTE timecode instead of ticks can't be read, so re-export those with standard timing.

How to find the BPM of a song

1. Click Choose an audio file, or drag the track onto the box.
2. Wait for the result. Longer tracks are checked in up to five 15-second sections, and the tempo those sections agree on most strongly wins.
3. If the song feels twice as fast or slow as the number, press Half time or Double time. For triplet feels, open Tempo correction for triplet feels and try ⅔× or 1.5×. Each button shows the tempo it would give before you press it.
4. For a second opinion, press Tap to the beat (or Space or T on your keyboard) in time with the music. The readout averages your last eight taps, the line under it shows the tempo over all of them and how even they are, and Reset taps (or R) clears only those taps.

Why the BPM can read as half or double

The detector listens for hits in two ranges, kick and bass in one and snares and hats in the other, then looks for the spacing that repeats most strongly. A 140 BPM trap beat and a 70 BPM half-time feel share most of that spacing, so either can win. Syncopated basslines and triplet patterns can throw it off by 2:3 in the same way. The correction buttons scale the current reading instead of re-running the analysis, so you can step between options until the number matches what you hear.

Confidence compares the winning tempo with every other tempo the detector considered: high means it stood well clear of the rest, and it counts for more when the 15-second sections agree with each other. It measures how clear the rhythm is, not whether the number matches the way you count it.

How to find the key of a song

1. Choose an audio file or drag one onto the box.
2. The tool checks the tuning first, then the key, using the loudest sections of the track.
3. Read the key and how confident the read is. The next closest key is listed under it, and a close call says so plainly.
4. Check the Tuning section for the offset in cents, the reference pitch and how far to pitch the track to reach A=440.

Detuned tracks and relative keys

Tape rips, sample-rate mismatches and deliberate pitch shifts can push a whole track a few cents sharp or flat. When the tuning read is clear, the key is measured against that shifted pitch instead of A=440, and the Tuning section tells you how many cents to pitch the track to bring it back to standard. The most common mix-up is between a major key and its relative minor, like C Major and A Minor, because they share the same notes. When those two are nearly tied, the tool checks the bassline to settle it. Keys are written with sharps, so D♭ Major appears as C♯ Major; the Key Equivalent Finder gives the flat name, the standard spelling and the Camelot code.

The file is mixed to mono at 8 kHz, and up to three of its loudest 15-second sections are analysed. The strength of each of the 12 notes there is correlated with the Krumhansl-Kessler major and minor key profiles in every key. The best fit is the key, and it counts as a strong match when its correlation leads the next closest key by 0.12 or more.

How to find a key’s other names

1. Pick the root note and Major or Minor, or type a key such as Ab major, G#m or 4B.
2. Read the other name at the top. Press it, or the arrows between the two names, to switch to that spelling.
3. Press the relative or parallel key to jump to it, or copy everything at once with Copy all key info.
4. Share the page address: it keeps the key you picked.

Why A♭ Major is written with flats

A key signature holds at most seven sharps or flats. A♭ Major has four flats. Spelled from G♯ it would need eight sharps, one of them the double sharp F, so sheet music uses A♭ Major even though both names sound identical in equal temperament. Some pairs, like D♭ Major (five flats) and C♯ Major (seven sharps), both have real key signatures, and the one with fewer accidentals is the usual choice. Scales are spelled with one letter per note, so A♭ Major contains D♭, never C♯. Minor keys use the natural minor scale, and the line under the notes gives the raised seventh of harmonic minor.

Camelot numbers the circle of fifths with C Major at 8B and A Minor at 8A. Open Key, used by Traktor, numbers the same wheel from C Major as 1d and A Minor as 1m, so its numbers are Camelot’s minus seven. Each step around either wheel is a fifth.

How to make a cover art mockup

1. Choose your cover art. Any image your browser can open will work, and square art fits best.
2. Optionally type the track title and artist name.
3. Check the three previews. On a phone, switch between them with Phone, Player card and Vinyl.
4. Press Download phone PNG, Download player card PNG or Download vinyl PNG for the ones you want.

Non-square artwork and long titles

Every art slot is square. If your image isn't, the tool crops it from the center to fill the slot instead of stretching it, so the edges of a wide or tall image get cut off; the page says which edges before you download. For a preview that matches your actual release, crop to 1:1 first with the Image Cropper and make sure nothing important sits near the edges. Titles and names that are too long for the layout are shortened with an ellipsis. If you'd rather have the artwork designed for you, see custom cover art.

Sizes

The phone PNG is 720 × 1480 pixels, the player card 1240 × 600 and the vinyl 1120 × 1120, drawn at twice the on-screen size so they stay crisp. The largest art slot is 480 pixels across, so artwork smaller than that is enlarged and looks a little soft. Every PNG carries a small rowen.work mark in the top-right corner.

How to make a visualizer video

1. Choose an audio file and press play to preview it.
2. Pick a style, an image shape (Landscape, Square or Vertical) and a color.
3. Open Background, Fine-tune animation, Shape and effects or Title and artist overlay for the finer settings.
4. Press Export video. The track restarts from 0:00 and records to the end, or press Stop and save to end early and keep what's recorded.

What to expect from the export

Recording happens in real time, so a three-minute song takes three minutes to capture, and you'll hear it play while it records. The picture is captured from the preview at 30 frames per second and about 6 Mbps, with the track's audio at 128 kbps. The file is WebM in most browsers, or MP4 where that's the only video format the browser can record; the page names the one you'll get next to the button. The video's pixel size follows how large the preview is on your screen, so a bigger window gives a higher-resolution file, and the exact size is shown before you start. Keep the tab in front while it records: browsers pause drawing in background tabs, and the video would freeze. Every export carries a rowen.work watermark. If Export video is greyed out, your browser can't record canvas video, so try Chrome or Edge.

Animation speed works by changing how much the sound analysis is smoothed between frames: slower settings smooth more, so the motion flows but trails the beat slightly. Beat flash and Reactive color follow the low end of the track, so they respond most to kicks and bass.

How to make a waveform image

1. Choose an audio file. The whole song is drawn, from start to end.
2. Pick a style (Bars, Mirror bars, Line, Filled, Blocks or Dots), an image shape (Landscape, Square, Vertical or Banner) and a color.
3. Open Background, Fine-tune shape or Title and artist text for the finer settings.
4. Under Download, pick PNG (at 1×, 2× or 3×) or SVG, then press Download PNG or Download SVG.

PNG or SVG: sizes and limits

Exports are 1600 pixels wide at 1×, so a 3× PNG is 4800 pixels wide. SVG keeps every bar, line and dot as a vector shape, which stays sharp at any size and can be recolored in a vector editor. A Transparent background works in both formats. The Cover image background only renders in PNG; an SVG export uses your solid background color in its place. Title and artist text in an SVG names its fonts rather than embedding them, so it can show in a fallback typeface on a computer that doesn't have them. Both formats include a rowen.work watermark.

Corner roundness, the line width in the Line style and the largest dot size are set in pixels of the finished image, so at 2× and 3× they look finer than in the preview. Everything else scales with the image.

How to make a lyric graphic

1. Type or paste the line into Lyric or quote. Line breaks you add are kept.
2. Pick a template and an image shape.
3. Open Text style, Position and alignment, Background or Credit line to fine-tune it. Show credit adds a line such as the song title and artist.
4. Pick PNG at 1×, 2× or 3×, or SVG, then press Download PNG or Download SVG.

The Spotify and Apple Music templates

These two swap the normal layout for a recreation of an app's lyrics view. Each line you type becomes a lyric line, the middle line is highlighted as the one playing, and a mini player shows the Song title and Artist you enter. If you upload a Cover image background, it also fills the album thumbnail. The icons are simple drawn shapes, not the apps' real logos. The templates always use their app's typeface; the Spotify Mix and SF Pro options under Font let you use those typefaces with the other templates. For these templates the PNG export is the safer choice, because the SVG version names its fonts instead of embedding them and the text can change typeface on another computer.

Sizes, fonts and backgrounds

Every image shape saves 1600 pixels wide at 1×, 3200 at 2× and 4800 at 3×; the height follows the shape. The text scales itself to the largest size that fits the frame at 100%. A Transparent background stays see-through in both PNG and SVG. A Cover image background is only kept in PNG; an SVG uses the solid background color in its place. Every download carries a rowen.work watermark.

1. Choose an audio file. The whole track starts out selected.
2. Drag across the waveform to select a section, drag the handles to fine-tune it, or type exact Start and End times.
3. Press Preview selection to hear the clip on a loop, with any fades and volume changes applied.
4. If you need them, open Fade in and out (up to 5 seconds each) or Volume (gain of ±12 dB and Normalize peak).
5. Press Download WAV. To save a WebM or change the name first, open File type and name.

WAV is written straight from the edited audio as 16-bit PCM at the file's own sample rate and channel count, and downloads as soon as it's rendered, so pick it for anything going back into a DAW. WebM is compressed, which makes it much smaller, but the browser has to play the clip through to record it, so the export takes as long as the clip. You can stop a WebM recording partway; nothing is downloaded if you do.

The preview plays the same render the download saves, so fades, gain and normalization are audible before you export. Play full track plays the original file, which is handy for finding the part you want. If you need an MP3, trim to WAV here and then run it through the File Converter.

1. Choose an MP3. Existing tags and artwork fill the form automatically.
2. Edit the title, artist and album. Open More track details for album artist, year, genre, composer, track and disc numbers, BPM and a comment. Genre offers the standard ID3 genre list, or Custom… for your own.
3. Press Add cover (or Replace cover) to set artwork, or Remove cover to clear it. Detect BPM fills in the tempo from the audio.
4. Set the file name and press Download MP3. Revert to original tags puts the form back to what the file had.

The editor removes any old ID3v2 tag at the start of the file and any ID3v1 tag at the end, then writes one new tag, so players don't find two conflicting sets of details. The audio itself isn't re-encoded. Only the fields on this form are written, so tags it doesn't show, such as embedded lyrics or an ISRC, won't carry over, and empty fields are left out. When a file has tags like that, the Download section lists them before you download. Uploaded cover images are saved as JPEG and scaled so the longest side is no more than 1000 pixels; a cover that was already in the file is kept as it is. Text is stored as UTF-8, so accented and non-Latin names come through.

The editor works with MP3 files only. It reads ID3v2.3 and ID3v2.4 tags, and fills the form from an older ID3v1 tag when that's all a file has. Tags in the rare ID3v2.2 format, or with an extended header, can't be read; the Download section says so, and saving replaces them.

1. Choose audio files. You can select several at once and add more later with Add clips.
2. Drag the grip on a clip to reorder it, or focus the grip and press the up and down arrow keys. Press Edit on a clip to set its gain, fade in, fade out or mute.
3. Pick how clips join: No gap, Silence gap (up to 5 seconds) or Crossfade (0.1 to 3 seconds).
4. Press play to hear the mix. Normalize peak is on by default; open Output level to turn it off and set Master volume yourself.
5. Press Download WAV. To save a WebM or change the name first, open File type and name.

In Crossfade mode each clip starts before the previous one ends, overlapping by the length you set, and the two ends fade out and in across the overlap. A crossfade can't be longer than either clip it joins. A muted clip keeps its place and length in the mix, so it plays as silence.

Normalize peak is on by default: after mixing, it scales the whole file so the loudest moment lands just under full scale, which also stops an overlap from clipping. Because it sets the final level, Master volume only appears when Normalize peak is off. The mix is rendered at the highest sample rate among the clips, with as many channels as the widest clip, and the WAV is 16-bit PCM. The preview plays the same render the WAV export saves. WebM export plays the mix through in real time to record it, so it takes as long as the track; you can stop it partway, and nothing is downloaded if you do.

1. Choose files or drag them onto the page. If you pick From and To at the top first, new files start on that conversion.
2. On each file’s row, press the format under Convert to and choose what that file becomes. The gear opens that file’s settings (quality, bitrate, trimming, video size and so on), with a button to use them for the other files too.
3. Press Convert. Files convert one at a time; each gets a Download button, and Download all (ZIP) appears once two or more are done. With two or more pictures in the list, Combine pictures into one PDF makes a single PDF instead.

What downloads, and when. Your browser already decodes common audio and images, so a WAV to MP3 or a PNG to JPEG downloads nothing. Other conversions need an engine, downloaded the first time a file needs it and then kept by your browser: the media engine (FFmpeg, 32 MB) for video and for audio formats like FLAC, M4A, OGG and WMA; the image engine (ImageMagick, 15 MB) for HEIC, TIFF, PSD, camera RAW, AVIF and GIF output; the PDF reader (about 2 MB); the archive engine (7-Zip, about 2 MB) for 7Z and RAR; and a font engine (about 1 MB) for WOFF2. Your files are still never sent anywhere.

Video. MP4, MOV, MKV and FLV are written as H.264 with AAC sound, WebM as VP8 with Opus, AVI as MPEG-4 with MP3, WMV as Windows Media, MPG as MPEG-2 and OGV as Theora. With Video quality on Best, a stream the new format can hold is copied untouched: an MKV of H.264 becomes an MP4 in seconds and loses nothing. Trimming or resizing always re-encodes.

Sample rates. The browser’s audio engine resamples to your sound card’s rate unless told otherwise, so the file’s own rate is read from its header first and kept: a 44.1 kHz track stays 44.1 kHz. MP3, WMA and AC3 can’t store more than 48 kHz, and Opus always plays at 48 kHz.

Pictures. Photos are turned the right way up from their EXIF data. Camera RAW is developed with the white balance the camera recorded. A PSD is converted from its flattened picture, so layers aren’t kept. JPEG and BMP can’t hold transparency, so transparent areas get the fill colour from the file’s settings.

Documents. Every document is read into headings, paragraphs, lists, tables and pictures, and every output is written from that, which is why any document can become any other. PDFs are written with the standard PDF fonts, which cover Western European languages; other characters show as a question mark.

One file at a time. Convert works down the list in order, on purpose: decoding several large files at once is how a phone runs out of memory halfway through. Stop after this file ends the batch between files, and Cancel on a row stops an engine mid-file.

1. Choose a project file, or drag it onto the box.
2. Check the summary, then the plugin and sample lists under Inside the project, to see what the project needs.
3. Under What you can do with this project, pick the version or option you want. Only the tools that work for that format appear, and the (?) beside each one says what it does.
4. Press the download button. The new file is named after yours, like mysong (Live 11).als, and your original is never changed.

What “save for an older version” really does. For Ableton and FL Studio, it rewrites the version stamp the project carries so the older app agrees to open it. It doesn’t translate anything else. Devices, plugins and features the older version doesn’t have are dropped or show as missing when it loads, so keep your original and check the result. A damaged FL Studio project isn’t re-saved, because rewriting it would hide the damage.

What MIDI export carries. Notes, their timing and velocity, track names and one tempo: the project’s starting tempo, used throughout. Audio, plugins, automation and mixer settings stay in the project. FL Studio exports one track per pattern with every pattern at bar 1, because the playlist isn’t read; LMMS leaves out Beat/Bassline steps.

Key. It’s read from the file when a Live 12 set or a MIDI key signature stores one; otherwise it’s estimated from the notes and labelled that way. Logic Pro, Cubase and Studio One projects use closed formats, so they can’t be read here.

1. Open a MIDI file, or choose a tempo, time signature and number of bars and press Start blank pattern.
2. Pick a track above the piano roll. Other tracks show faintly behind it.
3. Click the grid to add notes at the snap length, drag them to move, and drag their right edge to resize. Shift-click or drag a box to select several.
4. Use the velocity lane and Change notes to shape them, and press Play to hear them.
5. Press Download MIDI to take the file into your DAW, or Download WAV preview for an audio sketch.

Playback uses a small synth made in the browser: piano, electric keys, pluck, bass, pad and a basic drum kit for channel 10. It's there so you can check timing and notes, not to sound like your plugins, so load the downloaded MIDI into your DAW for the real sound. The editor keeps notes, the tempo map, the opening time signature, track names and instruments. Controller data such as sustain pedal, pitch bend and mod wheel is not shown and is left out of the saved file; when a file you open has any, the page lists it next to Download MIDI. Files timed in SMPTE timecode instead of ticks can't be opened; re-export them with standard timing.

1. Choose a WAV or FLAC bounce of the finished track.
2. Read What to check first: each Problem or Check says what was measured and what to listen for.
3. Look at Tonal balance against the guide closest to your genre, and at Stereo and mono for bands that cancel.
4. Press Play and switch between Phone, Laptop, Car and the other speakers, with Mono on and off.
5. Optionally add a released song you like under Compare with a reference and switch between them at matched loudness.

Loudness uses the standard K-weighting filter calculated for your file's own sample rate, 400 ms gating blocks and the -70 LUFS and -10 LU gates, and it reads the EBU Tech 3341 sine test signals it was checked against within 0.1 LU. The streaming table uses each service's published default target: -14 LUFS for Spotify, YouTube and Tidal, -16 for Apple Music. The tonal guides are hand-shaped curves, not averages of chart hits, so treat a difference as a prompt to listen rather than a fault. The technical score out of 100 adds up those measurements (20 points for streaming loudness, 15 each for true-peak headroom, dynamics, tonal balance and mono, 10 for clipping, 5 each for DC offset and loudness range) and shows every part. A quiet, dynamic or genre-specific master can score lower and still be right, and no number can say a mix sounds good. The time to full energy is a measurement of when the track gets loud, not a prediction of skips or plays.

1. Choose or drop an MP3, WAV, FLAC, OGG, Opus, AIFF or M4A file.
2. Read the summary at the top: format, length, file size, sample rate and channels. Codec, bit depth, bitrate and encoder are under Audio.
3. For WAV, FLAC and AIFF, check the result under Lossy history check.
4. Scroll down for tags, cover art and technical details, or press Copy details to paste them somewhere else.

For MP3 the tool tells constant from variable bitrate by reading the Xing, Info or VBRI header, and works out the average by scanning the frames. It reads tags from ID3v1 and all three ID3v2 versions, RIFF INFO, broadcast WAV, FLAC and Ogg comments, AIFF text chunks and M4A metadata.

The lossy history check: MP3 and AAC encoders cut off the highest frequencies to save space: around 16 kHz at 128 kbps, and 19 to 20 kHz at 256 to 320 kbps. Converting that MP3 back to WAV or FLAC doesn't bring those frequencies back, so the file keeps a flat shelf where the top end should be. The check averages the spectrum across the file and looks for that shelf. A shelf is a strong hint that the file had a lossy past. It can't prove it, though: a dark mix, a synth with a low-pass filter or a mastering filter can cut the highs in the same way, and a file with a full spectrum could still have been through a high-bitrate encoder. For MP3, OGG and AAC files there's nothing to check, since those formats are lossy by design.

1. Choose an audio file up to 10 minutes long.
2. Switch on the effects you want and adjust them; Distortion and bitcrush and Fades open with a click. Press Play to listen while you work, and switch to Original to compare.
3. Check Peak after effects. If it says the result clips, lower the gain or switch on Stop clipping.
4. Pick WAV or MP3 under Export settings, check the line under the button, then press Download WAV or Download MP3.

Trim, reverse, loop and bitcrush happen first, on the audio itself. Speed, distortion, volume and fades come after, in that order, so fade times are measured on the finished result and a sped-up loop keeps its crossfade. Speed works like a record played faster or slower, so pitch moves with it; this tool doesn't change tempo and pitch separately. Stop clipping turns the whole result down just enough that the loudest peak lands at −1 dB, rather than compressing the loud parts, so the sound stays the same apart from the level. If you only need to cut a section out, the Audio Trimmer is quicker, and Normalize Audio matches a target level.

How to preview your thumbnail

1. Choose your thumbnail, or up to three versions of it.
2. Type the title and read the Title cut-off line. Channel, views, age and length are optional.
3. Pick a layout and a theme, and choose which videos sit around yours.
4. Press Shuffle position a few times and see how quickly you find yours.

What's real and what isn't

The layouts copy the proportions of a video feed, not YouTube's exact styling. Trending and Search fill the feed with real public videos from YouTube, and pressing a category chip or searching inside the feed changes them. When those aren't available, or you pick Placeholders, the videos around yours are generic tiles with made-up titles; My images uses thumbnails you've saved from channels in your niche, which stay on your device. The title cut-off is measured in the font and width on screen, so it can differ by a few characters from what a viewer's device shows.

This tool can't tell you which version will get more clicks. Only real viewers can: YouTube Studio's Test & compare shows up to three thumbnails to real viewers and reports which one got more watch time.

1. Choose a thumbnail. Optionally open Add more context and type the video’s title, what it’s about and who it’s for.
2. Press Rate thumbnail for the score and design feedback. Open any category for the full report.
3. Under Local checks, pick what it’s for and where you’ll upload it from, and fix anything the upload box says YouTube won’t take. A tool link next to each problem opens a copy of the image there.
4. Look at the real-size previews. Optionally box the text and the main subject for size and contrast checks. Add up to two more versions to compare them side by side.

The thumbnail score is an AI-assisted design assessment, not a prediction of views or clicks. When you press Rate thumbnail, a vision model reviews the image against a fixed rubric: composition and visual hierarchy (20 points), main subject clarity and separation (15), small-screen readability (15), colour, contrast and lighting (15), visual hook and storytelling (15), typography and text integration (10), and craft and image quality (10). These are design-rubric weights, not a statistically validated model. The model rates each category from 0 to 10 against written anchors; the points and the total are calculated by this page, a category that doesn’t apply (no text, for example) is left out and the rest scaled to 100, and measured small-size contrast caps the readability rating. It only sends the image you’re viewing and the text you typed, and only when you press the button.

The local checks do three jobs. First it compares the file with YouTube’s own help page for the thing you’re uploading (a video, a Short or a podcast playlist) and the device you upload from, and keeps limits that stop an upload apart from recommendations that don’t. Then it measures how the picture holds up small, and turns anything notable into at most three suggestions, each saying what it saw, where, why it might matter and what to try. Last, it shows the thumbnail at the sizes feeds really use, blurred like a scroll past, and in greyscale.

What’s measured, what’s estimated, and what isn’t checked

Measured exactly: resolution, shape, file size and file type against YouTube’s page, and the size and position of any box you draw or the text finder adds, at each preview width. Estimated from the pixels: contrast that survives at 168 px wide, sharpness (measured on edges only, after removing grain, so noise and oversharpening can’t raise it), roughly how much real detail an upscaled image holds, grain, clutter, detail under the duration badge, and text contrast when the box holds two clear tones. These are labelled Likely or Possible and can be wrong, most often on deliberate styles. Not checked by the local checks: faces, what the picture shows, whether it fits the title, or whether anyone will click; the AI review judges the design side of those, and nothing here predicts clicks. Dark, bright, black-and-white, minimal and textured designs aren’t marked down for being so, and a thumbnail with no text or no face isn’t either.

How the local technical grade works

The grade out of 100 adds up fixed parts, all shown under Detailed analysis: 20 for fitting the chosen target’s limits and recommendations, 35 for contrast measured on a 168 px copy (where grain averages out), 20 for crisp edges at full size, 15 for a clear duration-badge corner (videos only; a box you draw under the badge counts too), and 10 for how busy the tile is at 168 px. Colour and saturation aren’t scored. Heavy grain lowers the sharpness part, because it can’t be measured through the grain; grain never adds points. It isn’t a click, creativity or CTR prediction.

Sharpness uses a no-reference blur measure (Crété-Roffet and others, 2007) on edge pixels after a median filter, and grain is estimated with Immerkær’s method (1996). The detail estimate repeats the blur measure on smaller copies: an upscaled picture only turns sharp near the size it came from. Expect it to be off by up to about half a step. Colourfulness is Hasler and Süsstrunk’s metric (2003). Text contrast is the WCAG contrast ratio between the two tones Otsu’s method finds in the box, and 3:1 is used as a reference point for large text. The 9 px text height and the 5% margin are rules of thumb, and the badge, progress bar and margin on the previews are approximations of YouTube’s current feed, not a specification. GIFs are checked on their first frame.

YouTube’s limits

The limits follow YouTube Help’s “Add custom thumbnails on YouTube” page as read on 27 September 2026: for videos, 3840 × 2160 recommended with a minimum width of 640; for Shorts, 2160 × 3840 with a minimum height of 640; “image formats such as JPG or PNG”; 16:9 for videos, 9:16 for Shorts and 1:1 for podcast playlists; and a size limit of 2 MB (videos) or 10 MB (podcasts) from a phone and 50 MB from a computer. Custom Shorts thumbnails can only be added in YouTube Studio on a computer, and custom thumbnails need a verified account.

How to download a thumbnail

1. Copy the video's link from the address bar or the Share button.
2. Paste it in and press Get thumbnail.
3. Press Download JPG for the largest size, or open Other sizes for the smaller ones.

Why a size can be missing

YouTube only makes the max resolution and standard sizes when the uploaded thumbnail or video was big enough. When a size doesn't exist, YouTube returns a small grey placeholder instead, and this tool hides it rather than offering you a 120 × 90 grey box. The standard and high sizes are 4:3, so they have black bars above and below a 16:9 picture. The images come straight from YouTube's servers to your browser. If your browser blocks the direct download, the image opens in a new tab for you to save. A thumbnail belongs to whoever made it, so don't reupload someone else's as your own.

How to write a title with it

1. Pick the video type.
2. Type the topic, or the artist for a type beat. Beat details (name, BPM, key, producer, [FREE]) are optional.
3. Press Shuffle order to see them in a different order, and copy the ones you like.
4. Edit the best one so it says what's actually in the video.

Templates, not AI

These titles come from a fixed list of formulas written by hand, filled in with exactly what you typed. That makes them predictable: "How to {topic}" needs a topic like "mix vocals", while "{n} best {topic}" reads better with "studio headphones". If one comes out clumsy, that's the template meeting your wording, so edit it. YouTube allows 100 characters, but search results usually cut titles at around 60 to 70 and phones cut them sooner, so put the important words first. Try the result in the YouTube Thumbnail Previewer to see where it gets cut next to your thumbnail.

How to build a description

1. Write a summary that says what the video is in its first sentence.
2. Open the sections you need (links, timestamps, credits, licensing, socials, hashtags) and leave the rest closed.
3. Read any warnings under the counter, especially about chapters. Make small changes straight in the finished description if you like.
4. Press Copy description and paste it into YouTube Studio.

Chapters, hashtags and the 150 characters

YouTube only turns timestamps into chapters when the first one is 0:00, there are at least three, and each chapter lasts at least 10 seconds. This tool warns you when those aren't met rather than inventing a 0:00 line. Only the first three hashtags go in, because those are the ones YouTube shows above the title. The "…more" cut is about 150 characters but moves with screen size, so treat the preview as a guide. Nothing here is written for you by AI; it only arranges what you typed.

Once you edit the finished description, it stops following the fields, so nothing you type in them overwrites your changes. Rebuild from the fields replaces your version with what the fields make now, and Bring back my edits restores it until you edit again.

How to make a tag list

1. Type the main topic and any keywords.
2. For beats, turn on Type beat tags and list the artists.
3. Remove any tag that isn't really about your video with its ×.
4. Press Copy tags and paste them into the Tags box in YouTube Studio.

How much tags matter, and how the count works

YouTube says tags play a small role in discovery and matter most for words people misspell, so the title, thumbnail and description do more work. Tags are still useful for type beats, where people search for the exact phrase. The counter follows YouTube Studio: each tag's characters, a comma between tags, and two extra for any tag with a space in it, because Studio wraps those in quotes. Tags that don't fit are listed separately under Don't fit, and Copy tags copies only the included ones. These tags are made from what you typed with fixed rules; there's no AI and no search-volume data behind them.

How to use an idea

1. Pick your niche and read through the concepts. Open one for its layout, colour, face and contrast notes.
2. Choose one that fits what actually happens in your video.
3. Make it, then check it at feed size in the YouTube Thumbnail Rater.
4. Drop it into the YouTube Thumbnail Previewer next to others from your niche.

Why the list is short

Each niche has six or seven ideas, written and checked by hand rather than generated on the spot, so Shuffle order reorders them instead of inventing new ones. They're written descriptions, not pictures: you make the thumbnail yourself. They're starting points, not rules. A thumbnail works when it matches the title and shows something the video delivers, and no list can know what that is for your video.

Duration badge

Almost every feed puts the video length in the bottom-right corner. A face, a word or a number there gets covered, so keep that corner for background.

Watch-progress bar

Anyone who has watched part of the video sees a red bar along the bottom edge. It’s thin, but text that touches the bottom gets underlined in red.

Hover buttons

On a computer, Watch later and Add to queue appear in the top-right corner while the pointer is over the thumbnail. They only show on hover, so this matters less than the badge.

Edge margin

A 5% margin keeps text off the rounded corners and away from the badge. It’s a guide, not a rule; a big face can break it on purpose.

Square and 4:5 crops

YouTube shows video thumbnails at 16:9, but the same art often ends up as a square or a 4:5 promo post. Anything outside those frames gets cut when that happens.

Center focus

Thirds lines and the middle of the frame, the places the eye lands first. Use them to check that the subject still reads when the thumbnail is 168 pixels wide.

How the guides are drawn

The guides are drawn as fractions of the frame, so they fit any 16:9 image. The badge and button sizes approximate how YouTube draws them on a desktop feed today; they aren’t a published spec, and YouTube changes its layout from time to time. The badge rectangle is the same one the YouTube Thumbnail Rater checks against. A thumbnail that isn’t 16:9 is shown the way a feed would frame it, with bars. Your image is read in the browser and never uploaded.

How to preview cover art

1. Drop your cover onto the box or choose the file.
2. Type the title and artist the way they’ll appear on the release.
3. Look at Small sizes first: that’s where most covers lose their text.
4. Press Check size and file to see whether the file meets common store requirements.

Sizes and what’s generic

Every scene is drawn at 1× CSS pixels, so a 48-pixel row is 48 pixels on your screen (on a high-density phone, that’s also how big it looks there). The layouts are generic players, not copies of Spotify, Apple Music or any other app, and the surrounding covers are plain placeholders. A cover that isn’t square is centered and cropped the way most apps do it. Nothing is uploaded.

Size

Spotify accepts square art from 640 to 10,000 pixels a side and asks you not to upscale. Apple Music asks for at least 1400 × 1400 and recommends 3000 × 3000 or larger. Most distributors ask for 3000 × 3000, so that’s the size to deliver. Enlarging a small image makes the file bigger, not sharper.

Shape

Both stores want exactly 1:1. A cover that’s a few pixels off square can be refused, so crop it rather than stretch it.

File type, colour and rotation

JPG and PNG are accepted everywhere; Spotify also lists TIFF. Colour should be RGB (Spotify asks for sRGB). A JPG saved in CMYK for print looks wrong on screens and is often refused. Some phones save photos sideways and add a rotation tag; Spotify says it doesn’t read orientation data, so that cover would show sideways.

Transparency

Transparent areas have no colour of their own, so each app fills them differently. Flatten the cover onto a solid background before you export.

The formula

One beat (a quarter note) lasts 60,000 ÷ BPM milliseconds, so at 120 BPM a quarter note is 500 ms. Each note value halves the one before it. A dotted note is 1.5 times as long, a triplet two thirds as long. Hz is 1000 ÷ milliseconds, and samples are milliseconds × 44.1 or × 48 for the two common sample rates. The pre-delays are a 1/64 and a 1/128 note; the decay is the note value minus the pre-delay, so the tail ends on the grid.

Equal temperament

Each semitone is the twelfth root of two (about 1.0595) times the one below it, and an octave doubles the frequency. So a note n semitones above A4 is A4 × 2^(n÷12). A cent is a hundredth of a semitone; the nearest-note readout rounds to the closest semitone and shows the difference in cents. MIDI note numbers use middle C as C4 = 60.

Ratios and semitones

The speed factor is new BPM ÷ original BPM, and the new length is the old length divided by it. Without pitch correction, the pitch moves 12 × log₂(speed factor) semitones, so doubling the speed is exactly +12, one octave. The in-tune tempos are the original BPM × 2^(n÷12) for whole numbers of semitones.