Skip to content

Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.

fix: pass --language auto to whisper.cpp for language auto-detection

Gabriel Steinfc1336ecf4db
diff --git a/src/helpers/whisperServer.js b/src/helpers/whisperServer.js
index 18e7a5ec7..a9314a178 100644
--- a/src/helpers/whisperServer.js
+++ b/src/helpers/whisperServer.js
@@ -233,9 +233,9 @@ class WhisperServerManager {
}
if (options.threads) args.push("--threads", String(options.threads));
- if (options.language && options.language !== "auto") {
- args.push("--language", options.language);
- }
+ // whisper.cpp defaults to English when --language is omitted;
+ // explicitly pass "auto" to enable language auto-detection
+ args.push("--language", options.language || "auto");
debugLogger.debug("Starting whisper-server", {
port: this.port,
@@ -404,13 +404,11 @@ class WhisperServerManager {
parts.push(finalBuffer);
parts.push("\r\n");
- if (language && language !== "auto") {
- parts.push(
- `--${boundary}\r\n` +
- `Content-Disposition: form-data; name="language"\r\n\r\n` +
- `${language}\r\n`
- );
- }
+ parts.push(
+ `--${boundary}\r\n` +
+ `Content-Disposition: form-data; name="language"\r\n\r\n` +
+ `${language || "auto"}\r\n`
+ );
// Add initial prompt for custom dictionary words
if (initialPrompt) {