[Config Support]: No processor handled the topic transcribe_audio

[jsapede]

Describe the problem you are having

having set up audio detectors flawlessly, i've tried to implement audio transcription but got errors !

i've setup global autio detectors and audio transcription to false but got audio transcription at true on a camera :

    audio_transcription:
      enabled: true

Version

0.17.1

Frigate config file

#################
# options audio #
#################
audio:
  enabled: true
  max_not_heard: 15
  min_volume: 350
  listen:
    - speech
    - bark
    - meow
    - scream
    - yell
    - fire_alarm
    - dog
    - cat
  filters:
    speech:
      threshold: 0.6
    scream:
      threshold: 0.65
    yell:
      threshold: 0.65


audio_transcription:
  enabled: false
  device: CPU
  model_size: large
  language: fr

Relevant Frigate log output

2026-05-18 16:33:56.741678914  [2026-05-18 16:33:56] frigate.embeddings.maintainer  ERROR   : No processor handled the topic transcribe_audio

Relevant go2rtc log output

no relevant error

Frigate stats

No response

Operating system

Debian

Install method

Docker Compose

docker-compose file or Docker CLI command

services:
  frigate:
    container_name: frigate
    privileged: true 
    restart: unless-stopped
    image: ghcr.io/blakeblackshear/frigate:0.17.1
    cap_add:
      - CAP_PERFMON
    shm_size: "512mb"
    devices:
      - /dev/dri/renderD128:/dev/dri/renderD128
      - /dev/dri/card0:/dev/dri/card0

    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /opt/frigate/config:/config
      - /opt/nfs/frigate/Model-cache:/config/model_cache
      #- /opt/nfs/media:/media/frigate
      - /opt/nfs/frigate/Medias:/media/frigate
      - type: tmpfs
        target: /tmp/cache
        tmpfs:
          size: 1G
    ports:
      - "5000:5000"
      - "8971:8971"
      - "1984:1984"
      - "8554:8554" # RTSP feeds
      - "8555:8555/tcp" # WebRTC over tcp
      - "8555:8555/udp" # WebRTC over udp
    environment:
      FRIGATE_RTSP_PASSWORD: "password"
      PLUS_API_KEY: XXXXX
      LIBVA_DRIVER_NAME: iHD
    labels:
      - "com.centurylinklabs.watchtower.enable=false"

Object Detector

OpenVino

Screenshots of the Frigate UI's System metrics pages

Any other information that may be helpful

No response

[inkeep-ai-assistant[bot]]

The error "No processor handled the topic transcribe_audio" occurs because you have audio_transcription disabled globally but enabled on a specific camera. (1)

For audio transcription to work, you must enable it at the global level. The camera-level enabled setting only controls whether that specific camera participates in transcription—it cannot override a globally disabled feature. (1)

To fix this, change your global config to:

audio_transcription:
  enabled: True
  device: CPU
  model_size: large
  language: fr

(1)

Then you can selectively disable it on cameras where you don't want transcription:

cameras:
  back_yard:
    ...
    audio_transcription:
      enabled: False

(1)

A few additional notes:

• Audio detection must also be enabled and configured for audio transcription features to work. (1)
• The device can be CPU or GPU. The whisper models can run on GPU but are only supported on CUDA hardware. (1)
• The large model setting uses whisper models which are more accurate but slower. Without a supported Nvidia GPU, generating transcriptions for longer speech events may take considerable time. (1)
• The small model uses sherpa-onnx which is fast and lightweight but always runs on CPU and is English-only for live transcription. (1)

Since you're using OpenVINO with Intel hardware (not CUDA), running the large whisper model on CPU may be slow for real-time transcription. You may want to consider using model_size: small for live transcription, though note it's English-only. Recorded speech events will still use the whisper model regardless of this setting. (1)

📚 Sources:

1. Audio Detectors

---

Have another question? Just tag @inkeep.

[hawkeye217]

The AI's response is incorrect - you should be able to only enable it for specific cameras. But there is a bug here where the post-processor is only registered when the global flag is also true.

I'll fix this for the next version of Frigate. For now, the workaround is to enable it at the global level. Because audio transcription is opt-in, it won't run on anything automatically, so enabling it at the global level won't cause any increased resource usage.

[jsapede]

yep, understood. its matter of privacy more than ressource.

w8ing for the next release.

having a remaining question, as frigate is used with homeassistant, would'nt it be possible to use already existing wyoming faster-whisper of homeassistant (instead having one more whisper instance ?). Moreover would it be possible to use online servies such as groq or openai ?

[hawkeye217]

its matter of privacy more than ressource.

You can just enable it at the global level, then explicitly disable it for the other cameras you don't want to use it for. That would achieve the same goal.

would'nt it be possible to use already existing wyoming faster-whisper of homeassistant (instead having one more whisper instance ?). Moreover would it be possible to use online servies such as groq or openai ?

Not at this time, but expanding the transcription features to remote providers is planned for a future version.

[hawkeye217]

This has been improved in #23238 and will be a part of Frigate 0.18.