how could I get ai_coustics works?
I only get the error from livekit.plugins.ai_coustics of: “Unable to authorize model use: Model use unauthorized”
seems there are a feature gate check, but we are on paid account
I’m running self-hosted agent but from doc / demo it should just work
Can you share your agent logs for where this happens?
19:54:17 voice.1 | {“message”: “voice session starting”, “level”: “INFO”, “name”: “pine.voice”, “conversation_id”: “178”, “room”: “voice_12_178_0010a8ef”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:17.258809+00:00”}
19:54:17 voice.1 | {“message”: “initializing process”, “level”: “INFO”, “name”: “livekit.agents”, “pid”: 86903, “timestamp”: “2026-04-09T11:54:17.265633+00:00”}
19:54:17 voice.1 | {“message”: “adaptive interruption detector initialized”, “level”: “INFO”, “name”: “livekit.agents”, “base_url”: “https://agent-gateway.livekit.cloud/v1”, “detection_interval”: 0.1, “audio_prefix_duration”: 1.0, “max_audio_duration”: 3, “min_frames”: 2, “threshold”: 0.5, “inference_timeout”: 0.7, “use_proxy”: true, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:17.310129+00:00”}
19:54:17 voice.1 | {“message”: “_run: transcript=None items=3 roles=[‘system’, ‘?’, ‘system’]”, “level”: “INFO”, “name”: “pine.voice”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:17.316288+00:00”}
19:54:17 voice.1 | {“message”: “_run: no transcript, has_instructions=True”, “level”: “INFO”, “name”: “pine.voice”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:17.316344+00:00”}
19:54:17 voice.1 | {“message”: “process initialized”, “level”: “INFO”, “name”: “livekit.agents”, “pid”: 86903, “elapsed_time”: 0.71, “timestamp”: “2026-04-09T11:54:17.979667+00:00”}
19:54:29 voice.1 | {“message”: “Processing failed: Unable to authorize model use: Model use unauthorized”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:29.711642+00:00”}
19:54:30 voice.1 | {“message”: “No VAD metadata found in frame.userdata[‘lk.aic-vad’] make sure that you are using noise_cancellation=audio_enhancement() on the audio input. This VAD plugin relies on its preprocessing.”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:30.261587+00:00”}
19:54:34 voice.1 | {“message”: “Processing failed: Unable to authorize model use: Model use unauthorized”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:34.761553+00:00”}
19:54:35 voice.1 | {“message”: “No VAD metadata found in frame.userdata[‘lk.aic-vad’] make sure that you are using noise_cancellation=audio_enhancement() on the audio input. This VAD plugin relies on its preprocessing.”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:35.263332+00:00”}
19:54:39 voice.1 | {“message”: “Processing failed: Unable to authorize model use: Model use unauthorized”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:39.811201+00:00”}
19:54:40 voice.1 | {“message”: “No VAD metadata found in frame.userdata[‘lk.aic-vad’] make sure that you are using noise_cancellation=audio_enhancement() on the audio input. This VAD plugin relies on its preprocessing.”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “pid”: 85522, “job_id”: “AJ_oyJXLVcVmkpr”, “room_id”: “RM_VU4NphJxxg6o”, “timestamp”: “2026-04-09T11:54:40.311865+00:00”}
with python sdk
livekit 1.1.5
livekit-agents 1.5.2
livekit-plugins-ai-coustics 0.2.7
I also see errors related to VAD plugin.
I shared a very basic agent that uses ai coustics in this thread:
Can you try with that basic agent to see if that works?
Thanks for the sample @darryncampbell 
Starting to wonder if it might be something on LiveKit’s side or can we as ai-coustics also do something about it.
@Dylan_Wang any luck with it?
@Pawel_Lach great to see you in the community, welcome 
There was an issue with our default starter agent where the noise cancellation settings would lead to error messages from ai-coustics. This was not an issue with ai-coustics or the integration with LiveKit, this was a problem with how the default starter agent applied noise cancellation, but since that agent is so ubiquitous we saw a number of issue reports.
The issue was fixed by Agent displays missing configuration for ai_coustics plugin · Issue #63 · livekit-examples/agent-starter-python · GitHub and since then the reports have stopped.
To be clear, I don’t think the above is the root cause for this topic, I’m just responding to your comment about maybe there is something wrong on the LiveKit side.
Amazing! Thanks let’s keep our eyes open on this in case the topic comes back again.
without luck I still getting this error
02:45:45 voice.1 | {“message”: “Processing failed: Unable to authorize model use: Model use unauthorized”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “room”: “voice_12_358_552743cf”, “pid”: 509, “job_id”: “AJ_naWzAMfQ7HyP”, “room_id”: “RM_T8fn9DZsvhWm”, “timestamp”: “2026-05-07T18:45:45.534819+00:00”}
02:45:46 voice.1 | {“message”: “No VAD metadata found in frame.userdata[‘lk.aic-vad’] make sure that you are using noise_cancellation=audio_enhancement() on the audio input. This VAD plugin relies on its preprocessing.”, “level”: “ERROR”, “name”: “livekit.plugins.ai_coustics”, “room”: “voice_12_358_552743cf”, “pid”: 509, “job_id”: “AJ_naWzAMfQ7HyP”, “room_id”: “RM_T8fn9DZsvhWm”, “timestamp”: “2026-05-07T18:45:46.086657+00:00”}
@Dylan_Wang If you were running it self-hosted, there was a workaround that but generally ai-coustics wasn’t supported out of the box. Soon we should get the update where you would be simply able to setup ai-coustics in the self-hosted environment by passing additional auth parameter.
I have tinkered around Today to create a solution for the newest release. Below I am sharing the md file that should have all of the necessary instructions to make it work but in short:
1. you create a separate file with regular aic SDK implementation
2. get ai-coustics sdk from our dev-platform ai-coustics dev portal
3. add prewarm to your agent entry file to download aic models
4. Pass the preloaded to room_io.AudioInputOptions(noise_cancellation=enhancer)
Check the attached md file for the intstructions.
Please disregard the previous post as new LiveKit ai-coustics plugin version has been released, and as of Today you can simply authenitcate ai-coustics SDK in the self-hosted environment by getting the ai-coustics SDK key from the website: https://ai-coustics.com/ and then authenticate it in the following way:
aic_audio_enhancement( model=EnhancerModel.QUAIL_L, auth=Auth.ai_coustics_api(license_key="API KEY HERE"), )
Cc: @Pawel_Lach, livekit-plugins-ai-coustics==0.2.13 shipped today on PyPI (2026-05-11). That’s the release Pawel’s post above is referring to; 0.2.7 from your earlier comment predates the new auth surface by several versions.
uv add "livekit-plugins-ai-coustics~=0.2.13"
Then plug in his block with the key from https://ai-coustics.com. Once the Unable to authorize error clears, marking Pawel’s post as the solution will help the next person hitting this find the version cutover fast.
Thanks @Muhammad_Usman_Bashir Indeed, we have also released Quail VF 2.1 model meanwhile, but that’s a bit different topic.