What is binaural-studio?

Binaural Studio is a browser-based tool for creating long headphone listening sessions. It combines generated binaural tones, your own music, adaptive sound textures, and spoken-word audio into a single mix you can record as one file. You use it with stereo headphones. The whole thing runs locally in your browser with no account, no server, and no uploads. The app centers on a ten-voice binaural synthesizer with seven preset states. Each preset shapes the tonal character of the generated layer. The synth works alongside a music deck that analyzes your tracks for musical key, tempo, and loudness before playback. That analysis lets the app tune the generated tones to match each song. A second deck handles spoken-word audio like audiobooks or podcasts, with its own volume, reverb, and timing controls. The two decks run independently, so spoken-word content never interferes with the music-driven analysis. Building a session involves adding music to the first deck, running the analysis, and optionally sorting tracks by key. You can assign state cues to individual tracks and adjust equalizer or crossfade settings. The second deck accepts common audio formats including MP3, M4A, WAV, FLAC, and Ogg, though the browser must support the actual codec inside the file. You can insert timed silences between spoken-word items or let the app place random gaps automatically. Recording happens in real time. The app captures the full master mix, including music, tones, textures, and spoken word, and saves it as an audio file. If your browser supports direct file access, it streams the recording to disk in one-second chunks. Otherwise it keeps everything in memory and downloads the file when the session ends. Nothing leaves your device. The app has no analytics and no server-side component. The project is plain HTML, CSS, and JavaScript with no dependencies. You can launch it directly from the hosted link or run it from a local server. The README notes some limitations: sessions do not persist across page reloads, recording cannot run faster than real time, and the key and tempo estimates can be unreliable on complex or sparse audio. The project is MIT licensed and welcomes contributions.