Implement Audio‑Input Features - Developer Guide

For developers requiring audio support, Infercom provides OpenAI’s Whisper large-v3 model, which enables real-time transcriptions and translations.

Whisper-Large-v3

• Model: Whisper-Large-v3
• Description: State-of-the-art automatic speech recognition (ASR) and translation model. Developed by OpenAI and trained on 5M+ hours of labeled audio. Excels in multilingual and zero-shot speech tasks across diverse domains.
• Model ID: Whisper-Large-v3
• Supported languages: Multilingual

Core capabilities

• Transcribes and translates extended audio inputs (up to 25 MB).
• Demonstrates high accuracy in speech recognition and translation tasks.
• Provides OpenAI-compatible endpoints for transcriptions and translations.

Request parameters

Parameter	Type	Description	Default	Endpoints
model	String	The ID of the model to use.	Required	transcriptions, translations
file	File	Audio file in FLAC, MP3, MP4, MPEG, MPGA, M4A, Ogg, WAV, or WebM format. File size limit: 25MB.	Required	transcriptions, translations
prompt	String	Prompt to influence transcription style or vocabulary. Example: “Please transcribe carefully, including pauses and hesitations.”	Optional	transcriptions, translations
response_format	String	Output format. Only json and text are honoured. See Response formats below.	json	transcriptions, translations
language	String	The language of the input audio. Using ISO-639-1 format (e.g., en) improves accuracy and latency.	Optional	transcriptions, translations
stream	Boolean	Enables streaming responses.	false	transcriptions, translations
stream_options	Object	Not supported. The endpoint rejects this parameter with HTTP 400 unknown_parameter.	-	Not accepted

Response formats

response_format accepts five values, but only two change the output.

Value	HTTP status	What you get
json	200	A JSON object: {"text": "..."}
text	200	The bare transcript string
verbose_json	200	The bare transcript string. Not a verbose object
srt	200	The bare transcript string. Not SRT subtitles
vtt	200	The bare transcript string. Not WebVTT subtitles

Caution

verbose_json, srt and vtt are accepted and return HTTP 200, but the endpoint ignores them and returns plain text. There are no timestamps, segments, or subtitle cues. Use json if you need a structured response.

Usage data

Non-streaming responses return the text field only. There is no usage object, so you cannot read token or duration counts from a transcription or translation response.

Note

stream_options: {"include_usage": true} does not work around this. The endpoint rejects the parameter with HTTP 400.

Example usage

Python (SambaNova SDK)

from sambanova import SambaNova
import base64

client = SambaNova(
    base_url="https://api.infercom.ai/v1",
    api_key="your-infercom-api-key",
)

audio_path="audio_path"
with open(audio_path, "rb") as audio_file:
   bin_audio = audio_file.read()

response = client.audio.transcriptions.create(
    model="Whisper-Large-v3",
    file=(audio_path,bin_audio),
)
print(str(response))

Python (OpenAI)

from openai import OpenAI
import base64

client = OpenAI(
    base_url="https://api.infercom.ai/v1",
    api_key="your-infercom-api-key",
)

audio_path="audio_path"
with open(audio_path, "rb") as audio_file:
   bin_audio = audio_file.read()

response = client.audio.transcriptions.create(
    model="Whisper-Large-v3",
    file=(audio_path,bin_audio),
)
print(str(response))

Javascript (SambaNova SDK)

import SambaNova from "sambanova";

const client = new SambaNova({
  baseURL: 'https://api.infercom.ai/v1',
  apiKey: 'your-infercom-api-key'
});

async function run() {
  const audioPath = 'audio_path';
  const binAudio = fs.createReadStream(audioPath);

  const response = await client.audio.transcriptions.create({
    model: 'Whisper-Large-v3',
    file: binAudio,
  });

  console.log('Full Response:');
  console.dir(response, { depth: null });
}

run().catch(console.error);