Replicate.com - Synthreo Builder

Purpose

The ReplicateCom node connects your workflow to AI models hosted on Replicate.com, enabling capabilities such as image generation, text processing, audio synthesis, document analysis, and more. It sends your workflow data to a selected model and returns the model’s output for downstream steps.

This node is useful when you need AI capabilities beyond what is built into the platform - for example, running a specialized open-source image model, a text-to-speech model, or a custom fine-tuned model that is deployed on Replicate.

---

What This Node Does

This node sends data from your workflow to AI models on Replicate.com and returns the AI-generated results. You can use it to:

• Generate images from text descriptions
• Process and enhance existing images
• Convert text to speech
• Analyze documents and extract information
• Create music or sound effects
• Transform data using advanced AI models

---

Inputs

• Dynamic data from prior nodes that will be passed to the chosen Replicate model.
• Specific inputs depend on the model and its parameters selected in the node.

---

Outputs

• AI-generated results (text, URLs to files, or structured data) from the selected model.
• Can be returned appended to original data or as result only, based on the Output Option setting.

Output Format (example):

{
  "replicate_response": "AI-generated result here (text or URL)"
}

For image generation models, the replicate_response value is typically a URL pointing to the generated file hosted on Replicate’s servers. If Download Output is enabled, the file is saved locally and the path is stored instead.

---

Parameters

Name	Type	Required	Default	Description
API Token (apikey)	Password	Yes	Empty	Your Replicate.com personal access token. Found in your Replicate account settings.
Model URL (modelApiUrl)	Text	Yes	Empty	The Replicate.com URL of the specific model to use (for example, https://replicate.com/stability-ai/sdxl).
Model Version (modelVersion)	Text	No	Empty	Exact model version hash for consistent, reproducible results. If left empty, the latest version is used.
Browse Models	Action button	No	-	Opens a searchable catalog of available Replicate models from within the node.
Model Parameters (modelParameters)	Visual editor	No	Empty	Model-specific input fields displayed after a model is selected. These vary by model.
Download Output (downloadOutput)	Toggle	No	Off	When enabled, automatically downloads generated files (images, audio, video) to local storage.
Download Column File Name (downloadColumnFileName)	Text	Required if download enabled	Empty	Naming pattern for downloaded files. Supports dynamic values from workflow data.
Download Directory (downloadDirectory)	Smart text	Required if download enabled	Empty	Folder path where downloaded files are saved. Supports dynamic values.
Output Option (outTransformId)	Dropdown	Yes	Original + appended result	Choose to append results to the existing data row or return only the result column.
Result Property Name (outColumnName)	Text	Yes	replicate_response	Name of the property that stores the AI results in the output data.

---

Output Option Values

Option	Description
Original + appended result	Keeps all upstream data and adds the model result as a new property. Use when downstream nodes need both the original inputs and the AI output.
Result only	Returns only the model result, discarding upstream fields. Use when only the generated content is needed.

---

Step-by-Step Configuration

1. Add Node and Connect: Place the ReplicateCom node on your workflow canvas and connect it to the node that provides the input data.
2. Authenticate: Enter your API Token in the Authentication section. Your Replicate personal access token is available at https://replicate.com/account/api-tokens.
3. Choose Model: In the Model section, either type a Model URL directly or click Browse Models to search and select a model from the catalog. You can browse available models at https://replicate.com/explore.
4. Pin a Version (Optional): Enter a specific Model Version hash if you need reproducible results and want to prevent the node from automatically using newer model versions.
5. Set Parameters: Configure the Model Parameters fields that appear after model selection. These fields are model-specific and may include inputs such as prompt text, image size, style options, or audio settings.
6. Configure Downloads (Optional): Toggle Download Output on if you want generated files saved locally. Provide a file naming pattern and a destination directory path.
7. Set Output Format: Choose the Output Option and set a descriptive Result Property Name so downstream nodes can reference the result clearly.
8. Test: Run a test with sample data to verify the model returns the expected output and that any file downloads are saved to the correct location.

---

Example Usage

Scenario 1 - Generating a product image from a description:

• Connect a data node that provides a product_description text field.
• Select a text-to-image model such as Stable Diffusion.
• Set the prompt model parameter to {{product_description}}.
• Enable Download Output and save images to a /product-images/ folder.
• Set Result Property Name to generated_image_path.

Scenario 2 - Converting text to speech for an IVR system:

• Connect a node that provides a message_text field.
• Select a text-to-speech model from the Browse Models catalog.
• Set the text model parameter to {{message_text}}.
• Enable Download Output to save the audio file locally.
• Pass the file path to a downstream node that uploads it to storage.

Scenario 3 - Running a custom fine-tuned classification model:

• Connect an upstream node that outputs a document_text field.
• Paste the full Replicate model URL into the Model URL field.
• Enter the specific Model Version hash to pin the deployment.
• Set model parameters as required by the model’s input schema.
• Use Result only output mode and name the column classification_result.

---

Troubleshooting

Issue	Likely Cause	Resolution
401 Unauthorized error	Invalid or missing API token	Verify the token in your Replicate account settings and re-enter it.
Model not found	Model URL is incorrect or the model has been removed	Use Browse Models to locate the current URL, or check the model page on replicate.com.
Output URL expires before download	Replicate-hosted file URLs expire after a short time	Enable Download Output to save files locally immediately after generation.
Parameter fields not appearing	Model URL has not been resolved yet	Ensure the Model URL is entered and press Enter or tab out of the field to trigger model metadata loading.
Download fails	Download Directory does not exist or is not writable	Verify the directory path exists and the workflow has write permissions.
Results are inconsistent between runs	No Model Version pinned	Enter a specific Model Version hash to ensure the same model weights are used every time.

---

Best Practices

• Pin model versions for production workflows. Replicate models are updated regularly, and a new version may change output behavior. Pinning a version ensures consistent results.
• Enable Download Output for any model that produces image, audio, or video files. Replicate-hosted output URLs are temporary and will expire.
• Use descriptive Result Property Names such as generated_image_url or speech_audio_path to make downstream node configuration easier to read and maintain.
• Test with small inputs first before running large batches, as Replicate API calls consume credits. Confirm the model output format matches your expectations before scaling up.
• Store API tokens securely. Do not hardcode them in body fields or pass them as visible text through the workflow. Always use the dedicated API Token credential field.

---

Related Nodes

• HTTP Client - for calling Replicate’s REST API directly if you need request-level control not provided by this node.
• Custom Script - for transforming or formatting input data before passing it to model parameters.
• Output Data - for surfacing generated images or text results to end users.