Skip to main content
Dolva accepts audio files uploaded as multipart/form-data in the audio field. For best results, your audio should meet the quality and format guidelines described on this page. Poor audio quality — excessive background noise, very low bitrate, or very short clips — can reduce signal accuracy. Dolva works with common audio formats. The following are recommended for best compatibility and analysis quality:
FormatExtensionNotes
WAV (PCM).wavBest quality; uncompressed
MP3.mp3Widely supported; moderate compression
M4A / AAC.m4aGood quality; common on mobile devices
FLAC.flacLossless compression; large file size
OGG Vorbis.oggOpen format; good compression
When possible, use WAV or FLAC to preserve the full acoustic signal. Heavily compressed formats like low-bitrate MP3 can reduce analysis accuracy.

​
Recording Guidelines

For accurate cognitive and emotion signal extraction, follow these recording best practices:
  • Minimize background noise — Record in a quiet environment. Loud background noise (traffic, music, crowd) reduces signal quality.
  • Use a close microphone — Headset or phone microphone held close to the speaker produces better results than far-field recording.
  • Avoid excessive clipping — Recording volume should be high enough to be clear but not so high that the audio distorts.
  • Include full utterances — Clips of at least a few seconds that contain natural speech yield the most reliable signals.

​
File Size

Keep audio files to a reasonable size for upload performance. For most conversational recordings, a few minutes of audio is sufficient. Very long files (e.g., several hours) should be split into segments before uploading.

​
Mono vs. Stereo

Dolva accepts both mono and stereo audio. If you have a stereo file (e.g., two speakers on separate channels), consider whether you want to analyze the full mix or extract individual channels for per-speaker analysis.

​
Language and Accent

Dolva’s acoustic models are language-agnostic — they analyze audio signal properties rather than words, so they work across languages and accents without additional configuration.
If you’re unsure whether your audio format is supported, test with a short clip first using the /v1/analyze/cognitive or /v1/analyze/emotion endpoint. A 422 Unprocessable Entity response may indicate a format issue.