AI Audio Processing and Transcription with Replicate - n8n Workflow

• Developers looking for production-ready n8n templates for AI model integration.


• Automation specialists needing to implement asynchronous API polling logic in n8n.


• Users who require automated speech-to-text transcription or audio analysis.


• Anyone needing a reliable n8n workflow to handle large, time-consuming AI generation tasks.

Generating complex AI outputs, such as high-fidelity audio or transcriptions, often requires an asynchronous process where the initial request only returns a prediction ID, requiring subsequent polling to retrieve the final result. This dedicated n8n workflow solves this complexity by providing a seamless, robust, and production-ready solution for interfacing with the Replicate API's notdaniel/voxtral-small-24b-2507 model.

This specific n8n workflow handles everything: securely configuring your API token, setting audio parameters (like the input audio URL and language), initiating the AI generation, and implementing a smart polling loop with built-in delays and success/failure branching. Using this n8n template ensures that your automations are resilient, reliable, and minimize API overhead by only checking status when necessary, making it an excellent example of advanced custom code scripting within n8n.

The entire process of this n8n workflow is structured around initiating an asynchronous task and monitoring its completion.

1. Start and Setup: The n8n trigger, a Manual Trigger node, starts the execution. It immediately passes control to the Set API Token n8n node, where the Replicate API key is securely stored as a variable, followed by the Set Audio Parameters n8n node, which defines the input audio file URL and desired parameters (like language 'en').


2. Request Initiation: The Create Audio Prediction n8n node sends a POST request to Replicate, containing the model version and input parameters. This request returns a unique prediction ID.


3. Monitoring: The n8n workflow logs the request details and then enters a polling loop, starting with a 5-second wait (Wait 5s).


4. Status Check: The Check Status n8n node uses the prediction ID to query the Replicate API for the task's current state.


5. Conditional Routing: An Is Complete? n8n node checks if the status is 'succeeded'.

1. Final Output: Both success and error paths converge at the Display Result n8n node, which presents the structured response object.

1. Import the n8n Workflow: Copy the provided JSON data and import it directly into your n8n instance.


2. Configure Credentials: Locate the Set API Token n8n node (d1782e9b) and replace the placeholder value 'YOURREPLICATEAPITOKEN' with your actual Replicate API key. This is critical for the n8n workflow to function.


3. Define Inputs: Navigate to the Set Audio Parameters n8n node (a31dc721) and modify the audio URL to point to the audio file you wish to process, or adjust the language and maxnew_tokens settings.


4. Execute: Activate the n8n workflow and click the Manual Trigger n8n trigger to run a test execution. Monitor the execution path to ensure the polling logic completes successfully and returns the expected result.

Manual Trigger (n8n trigger):
Function: Initializes the n8n workflow manually for testing and deployment.
Key Configuration: None (standard starting n8n trigger).
Set API Token (Set n8n node):
Function: Temporarily stores the Replicate API key using a Set n8n node, ensuring the token is available for downstream HTTP authentication.
Key Configuration: Assignment of variable apitoken.
Set Audio Parameters (Set n8n node):
Function: Defines the input variables required by the Voxtral AI model, including the audio URL, language, and token limit.
Key Configuration: Assignments for audio (the source file URL), language, and maxnew_tokens.
Create Audio Prediction (HTTP Request n8n node):
Function: Initiates the asynchronous audio generation task with Replicate, passing required parameters and authentication via headers.
Key Configuration: URL set to https://api.replicate.com/v1/predictions. Uses a Bearer token authorization header. Sends a JSON body defining the model version and input parameters.
Check Status (HTTP Request n8n node):
Function: Polls the Replicate API using the prediction ID received from the previous step to check the task status.
Key Configuration: URL dynamically built using expression: =https://api.replicate.com/v1/predictions/{{ $('Create Audio Prediction').item.json.id }}.
Is Complete? (IF n8n node):
Function: Directs the flow based on the result of the status check. If the status is 'succeeded', it exits the polling loop.
Key Configuration: Condition checks if status equals 'succeeded'.
Wait 10s (Wait n8n node):
Function: Implements a crucial delay in the polling loop, preventing API rate limit issues while the model processes the audio.
* Key Configuration: Amount set to 10 seconds. This retry logic ensures this n8n workflow is robust.