Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.
JavaScript aianthropiccross-platformgeminigroqlinuxmacosnvidiaopen-sourceopenaiparakeetspeech-to-texttranscribewhisperwindows
const crypto = require("crypto");
const fs = require("fs");
const path = require("path");
const { convertToMp3 } = require("./ffmpegUtils");
const { getSafeTempDir } = require("./safeTempDir");
// Containers every transcription provider we post to accepts as-is. Anything
// else the upload picker admits is re-encoded to MP3 before it leaves the app,
// so a .mov or .aiff upload never reaches a provider that would reject it.
const AUDIO_MIME_TYPES = {
mp3: "audio/mpeg",
wav: "audio/wav",
m4a: "audio/mp4",
webm: "audio/webm",
ogg: "audio/ogg",
oga: "audio/ogg",
flac: "audio/flac",
aac: "audio/aac",
opus: "audio/ogg",
};
function uploadExtension(filePath) {
return path.extname(filePath).slice(1).toLowerCase();
}
function providerContentType(filePath) {
return AUDIO_MIME_TYPES[uploadExtension(filePath)] || "audio/mpeg";
}
// Resolves to the path to upload plus a cleanup for any temp file it created.
async function prepareProviderUpload(sourcePath, { signal } = {}) {
if (Object.hasOwn(AUDIO_MIME_TYPES, uploadExtension(sourcePath))) {
return { path: sourcePath, cleanup() {} };
}
const mp3Path = path.join(
getSafeTempDir(),
`ow-upload-${Date.now()}-${crypto.randomBytes(4).toString("hex")}.mp3`
);
await convertToMp3(sourcePath, mp3Path, { signal });
return {
path: mp3Path,
cleanup() {
fs.rmSync(mp3Path, { force: true });
},
};
}
module.exports = { AUDIO_MIME_TYPES, providerContentType, prepareProviderUpload };