The toolbox covers the whole journey from a raw file to a finished karaoke video. You can extract the audio track from a video, trim and fade it in the online editor, change a song's pitch by up to twelve semitones in either direction while the tempo stays untouched, transpose a song into a comfortable key, remove or isolate vocals with optional AI separation, and build karaoke videos with timed lyrics — imported from an LRC or SRT file, or timed manually line by line.
The project's philosophy is browser-first. Everything that does not involve AI runs locally on your device through the Web Audio API: your audio, lyrics and videos are not uploaded anywhere, and nothing leaves the browser. The one exception is the optional AI vocal separation. It is clearly disclosed before it runs, it only starts when you explicitly trigger it, and the temporary files it creates are deleted automatically afterwards.
Rather than one overloaded app, the site is built as fourteen focused tools, one task per page. Each page does exactly one thing — extract, trim, transpose, separate, render — and links to its neighbours when the next step needs another tool. The whole interface is available in six languages: English, Français, Deutsch, Español, 日本語 and 한국어, with the same features in every language.
Exports are real files with honest limits. WAV download always works. MP3 download depends on whether your browser provides an encoder, and the page tells you which case applies. Video export uses the best encoder your browser offers — MP4 where it is supported, otherwise WebM, clearly labelled as such. Videos are limited to three minutes, 720p and 24 frames per second, and there is no watermark on anything you export.
One rule matters more than any feature: only upload music and lyrics that you own or are licensed to use. The tools process what you give them, but they grant no rights over the songs themselves. If you plan to publish a karaoke video or an instrumental in public, make sure your use of the underlying song is cleared. That part is your responsibility, and the privacy and copyright note on every page says so plainly.
Under the hood, the tools rest on standard web technology rather than closed plugins. Audio processing uses the Web Audio API, recordings are produced with MediaRecorder, and media files are parsed with the open-source Mediabunny library. The optional vocal separation is powered by Demucs, a research model for music source separation built on hybrid transformers. These are stated as facts, not endorsements — they are simply what the site runs on.
Karaoke Maker is a small, honest project. It was built around the things people actually need before singing: a clean instrumental, a key that fits the voice, lyrics that arrive on time. Where a limit exists it is written on the page, where a step is optional it is labelled as optional, and nothing pretends to be more than it is. That is the standard every tool here is held to.