Step 1: Upload or Record an Audio File
You can upload an existing file or record directly from your microphone.
Accepted formats: .wav, .mp3, .flac
Maximum length recommended: 60 seconds
Examples
🔹 english_meeting.wav – Short business meeting in English
🔹 french_news.wav – Excerpt from a French radio broadcast
🔹 spanish_podcast.wav – Segment from a Spanish-language podcast
Step 2: Start audio streaming
{}
Step 3: Configure the Task
Source Language
Chunks
1 60
0.1 5
0.1 10
Decoding
streaming_policy
“Wait-k: Higher accuracy, requires larger left context, higher latency”
”AlignAtt: Lower latency, suitable for production, predicts multiple tokens per chunk”
Detect hallucinations in the predicted tokens (default: True), works for both policies