Audio Input (Transcription)

LLM API supports multimodal models that accept audio as part of a chat completions request. You can send audio inline as base64-encoded data alongside text, and the model will transcribe or analyze it.

INFO

This feature uses the standard /v1/chat/completions endpoint — not /v1/audio/transcriptions. It is designed for multimodal models that process audio and text together, not dedicated STT models. For standalone transcription see Audio (TTS & STT).

Supported Models

Models that support audio input are marked with the Transcription capability on the models page.

Quick Start

# 1. Encode audio to base64
AUDIO_B64=$(base64 -i recording.mp3)

# 2. Send to chat completions
curl -X POST "https://api.llmapi.ai/v1/chat/completions" \
  -H "Authorization: Bearer $LLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"google-ai-studio/gemini-2.5-flash\",
    \"messages\": [
      {
        \"role\": \"user\",
        \"content\": [
          {
            \"type\": \"input_audio\",
            \"input_audio\": {
              \"data\": \"$AUDIO_B64\",
              \"format\": \"mp3\"
            }
          },
          {
            \"type\": \"text\",
            \"text\": \"Transcribe this audio\"
          }
        ]
      }
    ]
  }"

Request Format

The content field must be an array when including audio. Each content block has a type:

{
  "model": "google-ai-studio/gemini-2.5-flash",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "input_audio",
          "input_audio": {
            "data": "<base64-encoded audio>",
            "format": "mp3"
          }
        },
        {
          "type": "text",
          "text": "Transcribe this audio"
        }
      ]
    }
  ]
}

`input_audio` Block Fields

Field	Type	Description
`data`	string	Base64-encoded audio data (no `data:` URI prefix)
`format`	string	Audio format — see supported formats below

Supported Audio Formats

Format	Value
MP3	`mp3`
WAV	`wav`
OGG	`ogg`
M4A / AAC	`m4a` or `aac`
MP4	`mp4`
WebM	`webm`
FLAC	`flac`
Opus	`opus`

Python Example

import base64
from openai import OpenAI

client = OpenAI(
    base_url="https://api.llmapi.ai/v1",
    api_key="YOUR_KEY"
)

with open("recording.mp3", "rb") as f:
    audio_b64 = base64.b64encode(f.read()).decode()

response = client.chat.completions.create(
    model="google-ai-studio/gemini-2.5-flash",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "input_audio",
                    "input_audio": {
                        "data": audio_b64,
                        "format": "mp3",
                    },
                },
                {"type": "text", "text": "Transcribe this audio"},
            ],
        }
    ],
)

print(response.choices[0].message.content)

Combining Text and Audio

You can mix text and audio in the same message, and include conversation history with text-only turns:

{
  "model": "google-ai-studio/gemini-2.5-flash",
  "messages": [
    {
      "role": "system",
      "content": "You are an expert transcriptionist."
    },
    {
      "role": "user",
      "content": [
        {
          "type": "input_audio",
          "input_audio": { "data": "<base64>", "format": "wav" }
        },
        {
          "type": "text",
          "text": "Transcribe this and identify the speaker's language."
        }
      ]
    }
  ]
}

Billing

Audio input is billed separately from text tokens. The response usage object includes a breakdown:

{
  "usage": {
    "prompt_tokens": 16513,
    "completion_tokens": 2453,
    "prompt_tokens_details": {
      "audio_tokens": 16508,
      "cached_tokens": 0
    },
    "cost_usd_input": 0.0000015,
    "cost_usd_audio_input": 0.016508,
    "cost_usd_output": 0.006133
  }
}

Field	Description
`prompt_tokens_details.audio_tokens`	Tokens consumed by audio data
`cost_usd_audio_input`	Cost for audio tokens at the model's audio input price
`cost_usd_input`	Cost for text tokens only