Skip to content
orukSign up
Speech-to-textVoice agentsModelsPricingResearchAPI docsQuickstartSDKsMCP serverIntegrationsBenchmarksStatusLog in

Same transcript, two different deliveries

Hear two actors say “That is exactly what happened.” Compare identical transcriptions with their different saved emotion and speaking-style scores.

Listen to the recording

If a transcription system returns the same words twice, how can an application preserve the audible difference between the takes?

Take 1 · 4.20 seconds

Transcript: “That is exactly what happened.”

Stored Resonance 1 scores, take 1
OutputLabelScore
emotionshappy0.868
stylespassionate0.967

Download take 1 (WAV)

Take 2 · 3.04 seconds

Transcript: “That is exactly what happened.”

Stored Resonance 1 scores, take 2
OutputLabelScore
emotionsangry0.972
stylesirritated0.993

Download take 2 (WAV)

Reading the scores

The first take received happy and passionate. The second received angry and irritated. The performers differ as well as the intended delivery, so these two clips cannot isolate which acoustic cues caused the scores. Neither recording establishes whether the statement is true.

These acted English recordings were selected from the training split for demonstration. The scores describe audible delivery. Use a separate held-out set to measure accuracy.

Recording credits and licenses: CREMA-D. The public sample catalog includes each source file, audio hash and saved output.

Run this example

Shell · authenticated file analysis
curl --fail --location https://oruk.ai/samples/emotional/1005_TIE_HAP_XX.wav -o take.wav
curl --fail-with-body https://speech-api.oruk.ai/v1/audio/analysis \
  -H "Authorization: Bearer $ORUK_API_KEY" \
  -F file=@take.wav -F model=oruk-resonance

The example uses a key with speech API access. The table shows the saved run. Open the matching SDK quickstart.

Try it on your own audio.

The demo needs no account. The trial needs a card.