Create a Digital Human

10min

To create a digital human you will need a user and organisation. Depending on your organization's type and privileges, you will have access to various head visuals. See User﻿ for more information. 
Components
All Digital Humans are made up of the following required components:
Name
Alias
Face
Voice
Operating Mode
Prior to starting the creation process, it is important to consider what operating mode your Digital Human should use. This is directly related to the intended use case. 
Operating Modes
Digital Humans can operate in 5 distinct modes
Text-to-Video: Specify text for a given Digital Human, and a video will be generated with the digital human speaking the text. The output is an mp4. 
Open Conversation: Configure a prompt for a given Digital Human, and the Digital Human will be conversational. The output is a hosted, conversational Digital Human.
Knowledge base: Provide content and configure a prompt, and the Digital Human will be conversational about the content provided. The output is a hosted, conversational Digital Human.
Guided Conversation: Build a workflow, and the Digital Human will be able to guide you through the conversational workflow. The output is a hosted, conversational Digital Human 
Plugin mode: Leverage a webhook to connect any custom conversational engine or LLM to power the conversation of the Digital Human. The output is a hosted, conversational digital human .
Text-to-Video operating mode is only recommended for testing purposes. The Digital Human will be available to review how it speaks the text you provide it, in real-time. Text-To-Video is a feature available across all Operation Modes. See Generate a Digital Human Video﻿.
Digital Human Creation Process
All digital humans need an existing head visual and voice. To see your available head visuals, see List Faces﻿ and List Voices﻿ documentation.
Names and aliases are free text fields which are used for personalization. There are no restictions apart from being required fields. 
Use the POST /head/create- to create Digital Humans. Each operation mode requires a different request body (see Operating Mode Parameters﻿). 
doc_qa
ttt
oc
voiceflow
plugin
{ 
  "headVisualId": "<stringIdRetrievablefromListFaces>",
  "name": "testoc",
  "alias": "Open AI Salesperson",
  "languageSpeechRecognition": "en-US",
  "langCode": "en-US",
    "operationMode": "oc",
    "promptConfig": {
    "system_prompt": "Act as a salesleader called Carmen who wants to help her team perform better",
    "qa_prompt": {
      "introduction": "",
      "personality": "",
      "qa_instructions": ""
    },
    "intentions":[]
  },
  "ttsProvider": "audiostack",
  "ocProvider": "playground",
  "ttsVoice": "carmela",
  "greetings":"Hi there!",
  "phrases": ["Unith","Barcelona"],
  "customWords": {"Barcelona":"Barthelona","AI":"A eye"}
}
{ 
  "headVisualId": "<stringIdRetrievablefromListFaces>",
  "name": "testoc",
  "alias": "Open AI Salesperson",
  "languageSpeechRecognition": "en-US",
  "langCode": "en-US",
    "operationMode": "oc",
    "promptConfig": {
    "system_prompt": "Act as a salesleader called Carmen who wants to help her team perform better",
    "qa_prompt": {
      "introduction": "",
      "personality": "",
      "qa_instructions": ""
    },
    "intentions":[]
  },
  "ttsProvider": "audiostack",
  "ocProvider": "playground",
  "ttsVoice": "carmela",
  "greetings":"Hi there!",
  "phrases": ["Unith","Barcelona"],
  "customWords": {"Barcelona":"Barthelona","AI":"A eye"}
}
﻿
The payload can contain many additional properties of the Digital Human as described in the Configuration Parameters﻿ page. 
If you have access to multiple organsations, you will need to add  "orgId":"<orgId>" to the payload. 
Request Body Parameters
Parameter
Data Type
Description
headVisualId
string
Head Visual ID:  The unique identifier of the visual representation (the "look") chosen for the digital human, selected from a list of available head visuals.
orgId
string
Organization ID:  The unique identifier of your organization. This parameter is used to associate the digital human with your organization account.
suggestions
string
Used in "document-based question answering" and open dialogue;(operationMode=doc_qa, operationMode=oc.  Allows manual overriding or adding to automatically extracted suggestions from uploaded documents. Format as a JSON array string, e.g., ["example1", "example2"].
name
string
The system-generated name for the digital human, derived from the alias. It's generally recommended to leave this parameter untouched.
alias
string
The user-defined, short, and unique name for your digital human. This alias will be used to construct the public name of the digital human.
languageSpeechRecognition
string
Sets the language for speech recognition, using codes supported by Microsoft Azure Speech Services. Refer to [Microsoft Azure documentation]for a full list of supported languages.
language
string
Frontend Language:  Sets the language for the user interface and frontend elements of the digital human. To get the list of currently supported languages please use the endpoint:
GET/languages/all
ttsProvider
string
Text-to-Speech Provider:  Specifies the service used for text-to-speech generation.  Use the value "audiostack" as the primary wrapper service.
operationMode
string
Operation Mode: Defines the operational behavior of the digital human. Possible values are: "oc" (open conversation), "doc_qa" (document-based question answering), "ttt" (text-to-talk - video generation from text input) and "plugin"for custom plugin implementation
promptConfig
string
Used to customize the behavior of the digital human for "doc_qa" and "oc" operation modes.  See nested parameters below for details.
system_prompt (nested promptConfig)
array
 (For operationMode = oc) Sets the overall system prompt to guide the behavior and personality of the digital human in open conversation mode.
qa_prompt (nested promptConfig)
string
 (For operationMode = doc_qa) Contains settings to adjust the question-answering behavior in document-based conversation mode. See nested parameters below.
introduction (nested qa_prompt)
string
deprecated
personality (nested qa_prompt)
string
(For operationMode = doc_qa) Allows you to adjust the personality or tone of the digital human when answering questions based on documents.
qa_instruction (nested qa_prompt)
string
(operationMode = doc_qa) Provides specific instructions to guide the question-answering behavior. It is important to keep instructions related to response retrieval untouched.
intentions (nested promptConfig)
array
Custom "intentions" can be configured to be automatically included before long responses, thereby minimizing perceived latency.
ttsVoice
string
Text-to-Speech Voice:  Selects a specific voice for text-to-speech output. Choose from a wide range of voices across different providers. See [voice list documentation]for available voices.
greetings
string
Sets the initial greeting message that the digital human will use to start a conversation.
voiceflowApiKey
string
Required only when operationMode is set to "voiceflow".  Paste your Voiceflow conversation API key here to connect the digital human to your Voiceflow conversation flow.
isRandomSuggestions
boolean
Random Suggestions:  Determines whether suggestions are displayed in a fixed order (false) or in a randomized order (true). Defaults to true if not specified.
pluginOperationalModeConfig
object
Plugin Configuration:  Used when operationMode is set to "plugin".  Allows specifying plugin-specific configurations. Refer to the plugin documentation for detailed payload structure. 
customWords
object
Custom Words:  Allows defining custom pronunciations for specific words.  Provided as a key-value object (dictionary) where keys are the words and values are their custom pronunciations (e.g., {"unith": "iunit"}). Case-sensitive.
﻿
Nested promptConfig Parameters: (Only relevant for operationMode = doc_qa and oc)
JSON
"promptConfig": {
  "system_prompt": "string",
  "qa_prompt": {
    "introduction": "string",
    "personality": "string",
    "qa_instructions": "string"
  },
  "intentions": "string"
}
"promptConfig": {
  "system_prompt": "string",
  "qa_prompt": {
    "introduction": "string",
    "personality": "string",
    "qa_instructions": "string"
  },
  "intentions": "string"
}
﻿
Example requests  
Curl
curl -X 'POST' \
  'https://platform-api.unith.ai/head/create' \
  -H 'accept: application/json' \
  -H 'Authorization: Bearer abcasdlkjasdfljaldsfjñsldfkj' \
  -H 'Content-Type: application/json' \
  -d '{
  "headVisualId": "abc123abc123",
  "name": "ttt",
  "alias": "Repeater",
  "languageSpeechRecognition": "en-US",
  "langCode": "en-US",
  "ttsProvider": "audiostack",
  "operationMode": "ttt",
  "ocProvider": "playground",
  "ttsVoice": "coco"
}'
curl -X 'POST' \
  'https://platform-api.unith.ai/head/create' \
  -H 'accept: application/json' \
  -H 'Authorization: Bearer abcasdlkjasdfljaldsfjñsldfkj' \
  -H 'Content-Type: application/json' \
  -d '{
  "headVisualId": "abc123abc123",
  "name": "ttt",
  "alias": "Repeater",
  "languageSpeechRecognition": "en-US",
  "langCode": "en-US",
  "ttsProvider": "audiostack",
  "operationMode": "ttt",
  "ocProvider": "playground",
  "ttsVoice": "coco"
}'
﻿
Doc_qa Digital Humans need an additional step to be functional. This is described in Create a doc_qa Digital Human﻿ and related to uploading a knowledge document. 
Interact with a Digital Human
Upon creating a Digital Human with a single call to the head/create  endpoint, your Digital Human is now hosted by unith at chat.unith.ai/<generatedPath>
This path can be found in the  POST head/create response, or in the  GET head/{ID} response:
Public URL
{
  ...
  "publicId": "headId",
  "publicUrl": "https://chat.unith.ai/[org-id]/[head-id]?api_key=[org-api-key]",
  ...
}
{
  ...
  "publicId": "headId",
  "publicUrl": "https://chat.unith.ai/[org-id]/[head-id]?api_key=[org-api-key]",
  ...
}
﻿
Embed a Digital Human in your application
See Embed Digital Humans in your Application﻿﻿

Parameter	Data Type	Description
headVisualId	string	Head Visual ID: The unique identifier of the visual representation (the "look") chosen for the digital human, selected from a list of available head visuals.
orgId	string	Organization ID: The unique identifier of your organization. This parameter is used to associate the digital human with your organization account.
suggestions	string	Used in "document-based question answering" and open dialogue;(operationMode=doc_qa, operationMode=oc. Allows manual overriding or adding to automatically extracted suggestions from uploaded documents. Format as a JSON array string, e.g., ["example1", "example2"].
name	string	The system-generated name for the digital human, derived from the alias. It's generally recommended to leave this parameter untouched.
alias	string	The user-defined, short, and unique name for your digital human. This alias will be used to construct the public name of the digital human.
languageSpeechRecognition	string	Sets the language for speech recognition, using codes supported by Microsoft Azure Speech Services. Refer to [Microsoft Azure documentation]for a full list of supported languages.
language	string	Frontend Language: Sets the language for the user interface and frontend elements of the digital human. To get the list of currently supported languages please use the endpoint: GET/languages/all
ttsProvider	string	Text-to-Speech Provider: Specifies the service used for text-to-speech generation. Use the value "audiostack" as the primary wrapper service.
operationMode	string	Operation Mode: Defines the operational behavior of the digital human. Possible values are: "oc" (open conversation), "doc_qa" (document-based question answering), "ttt" (text-to-talk - video generation from text input) and "plugin"for custom plugin implementation
promptConfig	string	Used to customize the behavior of the digital human for "doc_qa" and "oc" operation modes. See nested parameters below for details.
system_prompt (nested promptConfig)	array	(For operationMode = oc) Sets the overall system prompt to guide the behavior and personality of the digital human in open conversation mode.
qa_prompt (nested promptConfig)	string	(For operationMode = doc_qa) Contains settings to adjust the question-answering behavior in document-based conversation mode. See nested parameters below.
introduction (nested qa_prompt)	string	deprecated
personality (nested qa_prompt)	string	(For operationMode = doc_qa) Allows you to adjust the personality or tone of the digital human when answering questions based on documents.
qa_instruction (nested qa_prompt)	string	(operationMode = doc_qa) Provides specific instructions to guide the question-answering behavior. It is important to keep instructions related to response retrieval untouched.
intentions (nested promptConfig)	array	Custom "intentions" can be configured to be automatically included before long responses, thereby minimizing perceived latency.
ttsVoice	string	Text-to-Speech Voice: Selects a specific voice for text-to-speech output. Choose from a wide range of voices across different providers. See [voice list documentation]for available voices.
greetings	string	Sets the initial greeting message that the digital human will use to start a conversation.
voiceflowApiKey	string	Required only when operationMode is set to "voiceflow". Paste your Voiceflow conversation API key here to connect the digital human to your Voiceflow conversation flow.
isRandomSuggestions	boolean	Random Suggestions: Determines whether suggestions are displayed in a fixed order (false) or in a randomized order (true). Defaults to true if not specified.
pluginOperationalModeConfig	object	Plugin Configuration: Used when operationMode is set to "plugin". Allows specifying plugin-specific configurations. Refer to the plugin documentation for detailed payload structure.
customWords	object	Custom Words: Allows defining custom pronunciations for specific words. Provided as a key-value object (dictionary) where keys are the words and values are their custom pronunciations (e.g., {"unith": "iunit"}). Case-sensitive.

Updated 13 Feb 2025

Did this page help you?

User

Update a Digital Human