Workloads

Transcription

Whisper-large and distil-whisper on bulk audio. Speaker diarisation, word timestamps, and language detection are optional.

python

job = cc.transcription.create(
    "call.wav",
    model="whisper-large-v3",
    timestamps=True,
)
print(job.wait().output)

For a directory of recordings, fan out with submit_many — results stream back as each file finishes.

python

from pathlib import Path

for result in cc.submit_many(
    "whisper_ane",
    [{"input_uri": u} for u in presigned_urls],
    priority="batch",
    max_spend_usd=25,
):
    save(result.task_id, result.output)