The Question Answering task (get_answer) treats an audio transcript as a spoken question and returns a direct answer instead of a summary. It is category-independent and works across every supported language.
When get_answer is TRUE, the transcript is treated as a question and only the answer (transcript_answer) is returned. All other post-processing options (summary, SOAP note, entity list, etc.) are ignored for that request, regardless of the selected use_category.

​
How it works

  1. The audio is transcribed in the input language (use_language_asr_input).
  2. The transcript is passed to the LLM, which detects whether a question or request is present and answers it directly and concisely.
  3. If no answerable question is present, the model briefly acknowledges what was said and invites the user to ask a question rather than inventing one.
  4. The answer is returned as transcript_answer in the response.

​
Answer language

By default the answer is produced in the language of the question (the input language). To force a different answer language, set use_language_data_extraction_output to the desired language code.
use_language_asr_inputuse_language_data_extraction_outputAnswer language
sw (Swahili)(not set)Swahili
sw (Swahili)en (English)English
en (English)(not set)English

​
Endpoint

POST https://infer.voice.intron.io/file/v1/upload/sync

​
Authentication

All requests require a Bearer token in the Authorization header: 
Authorization: Bearer YOUR_API_KEY

​
Request Body form-data

FieldTypeDescriptionRequiredOptionsDefault
audio_file_nameStringnon-unique file nameyes
audio_file_blobStringthe audio file containing the spoken questionyes
use_categoryStringcategory of the file (any category is valid)nofile_category_telehealth
use_language_asr_inputStringinput language code of the spoken questionnoen
use_language_data_extraction_outputStringlanguage code for the answer; defaults to the question languageno(question language)
get_answerStringtreat the transcript as a question and return only the answeryesTRUE | FALSEFALSE

​
Example: Swahili question, English answer

The example below sends a Swahili audio question using the file_category_general category and requests the answer in English by setting use_language_data_extraction_output to en. 
curl --location 'https://infer.voice.intron.io/file/v1/upload/sync' \
  --header 'Authorization: Bearer api-key' \
  --form 'audio_file_name="my_question"' \
  --form 'audio_file_blob=@"/C:/Users/aaaa/question.wav"' \
  --form 'use_category="file_category_general"' \
  --form 'use_language_asr_input="sw"' \
  --form 'use_language_data_extraction_output="en"' \
  --form 'get_answer="TRUE"'

​
Response

The transcript (audio_transcript) is returned in the question’s language, while the answer (transcript_answer) is returned in the requested output language.
The request can timeout with an http status code 503 and the file-id in 120 seconds. The file-id in the response can be used to get the file’s status with the Get File Status endpoint.
{
    "data": {
        "file_id": "12a9760f-b165-4404-91d0-a65d4cdt78fs",
        "processing_status": "FILE_TRANSCRIBED",
        "audio_file_name": "my_question",
        "audio_transcript": "Mji mkuu wa Kenya ni upi?",
        "processed_audio_duration_in_seconds": 6,
        "transcript_answer": "The capital city of Kenya is Nairobi."
    },
    "message": "file status found",
    "status": "Ok"
}

​
Next Steps

  • Use the Get File Status endpoint to retrieve results if the synchronous request times out.
  • See Supported Languages for the full list of input and output language codes.