Audio moderation API for voice and audio files.

Voice messages and audio files are transcribed and checked against the policy you already use for text, across 35+ harm categories.

Get startedTalk to an expert
Voice message · vm-7720@user_412
0:14
Transcript
0:02yeah I saw your stream last night and honestly
0:07if you show up again I know where you live
0:12so think about it
Threat · 0:07-0:11Same policy as text
eBaySupercellPolymarketKalshiStravaRobinhoodStockXPhantom

Trusted for voice at scale.

Read the benchmark
  • 33M+audio minutes moderated/month
  • 50+languages
  • 35+harm categories

Everything your text policy catches, spoken out loud.

All 35+ text categories apply to audio, including:

See every text category
#CategoryWhat it catches
01Harassment and hate speechInsultsTrollingHumiliationHatredRacismMisogyny
02ThreatsThreatsViolenceSelf-harm
03Scams and platform bypassScamsSpamLinksPlatform bypass
04Sexual contentSexually explicit contentSextingDating solicitation

Transcribe. Check. Act.

One policy, no separate setup.

  1. 01

    Transcribe

    Stream converts the audio to text.

    Transcript
    ...if you show up again I know where you live
  2. 02

    Check

    Your existing text rules check the transcript.

    ...if you show up again I know where you live
    Threat · high
    Same labels and severity as your text policy
  3. 03

    Act

    Your rules remove, warn, ban or send it to review.

    Rule: Threat ≥ high -> Remove
    RemoveWarnBanSend to review
    Applied automatically and logged for review

Abuse can't hide in a voice note.

Voice messages in DMs and group conversations are checked like text, with rules for removals and bans.

are you coming tonight?
0:09
0:14
Harassment · removed

Find the flagged moment in any recording.

Moderators see the flagged transcript segment with user history, so they review in seconds. Nora, your moderation agent, handles the clear cases first.

Explore the dashboard
Voice message · vm-7720Threat · high
0:14
Transcript
0:02yeah I saw your stream last night
0:07and if you show up again I know where you live
0:12so think about it
User history
HarassmentSep 28
WarnedSep 28
JoinedAug 2
Ban userDismiss

Private by default.

Audio is processed to produce a transcript and a decision. You control what is kept and for how long.

See compliance
  1. 01

    Audio files

    Configurable retention periods based on your company's needs.

  2. 02

    Transcripts

    Kept for 30 days within the dashboard for review and audit, then indefinitely queryable via API.

  3. 03

    Decisions and audit log

    Every action is logged with its reasoning to support DSA and Online Safety Act transparency reporting.

Moderate any audio, from any platform.

Send any audio file through one API call, from your own app or storage.

Read the docs
const res = await client.moderation.check({
  entity_type: "voice_message",
  entity_id: "vm-7720",
  entity_creator_id: "user-412",
  moderation_payload: {
    audio: ["https://cdn.app.com/vm-7720.m4a"],
  },
  config_key: "voice",
});

// res.recommended_action -> "flag"

Stream moderation FAQs.

More content types, same policy.

Moderation overview

Test it on your voice content.

Send a sample of voice messages or recordings and see the results at no cost.

Get startedTalk to an expert