Frequent Problems

Can't connect to the API

Or you get disconnected shortly after connecting.

1. Check your balance.
2. Soon: Check our API availability page for any ongoing incidents.
3. Check if any error messages are returned and contact Palabra.ai support.

Not recognizing speech

No transcription is returned when you speak.

1. Lower min_alignment_score to 0 first. If the issue disappears, gradually increase it until it starts missing words again. This is especially relevant for non-native speakers. The higher you can set the score, the better. Lower scores increase the probability of hallucinations (random phrases appearing), especially with poor audio quality.
2. If the audio is very noisy, set denoise to alpha. Note that on good-quality audio, this can decrease ASR accuracy. Only enable denoise if the signal is really bad. Another option is to enable pre_vad_denoise in the preprocessing section.
3. For non-primary languages: Ensure that asr_model is set to alpha, not alpha_turbo.
4. Move the microphone closer and speak as clearly as possible. This helps when you're a non-native speaker and your speech is frequently cut off due to a low alignment_score.

Returns words you didn't pronounce

1. Gradually increase min_alignment_score by 0.05 until it starts failing to recognize your speech (values above 0.4 are generally not recommended).
2. Follow steps 2, 3, and 4 from the section above.

Not translating

You receive a transcription, but no translated text or TTS output.

1. Change translation_model to another option: alpha, beta, or gamma.
2. Disable sentence_splitter.
3. Disable verification.

Not generating speech

You do not hear any translated speech.

1. Change tts_model from alpha to beta (or vice versa).
2. Disable voice_cloning and voice_timbre_detection, and set voice_id to default_low.