Audio

Convert text to lifelike speech and transcribe audio to text. OpenAI-compatible endpoints with support for Sarvam AI Indian language voices.

Convert text to lifelike speech and transcribe audio to text using OpenAI-compatible endpoints. Both /audio/speech and /audio/transcriptions are available with costs charged in INR from your wallet.

Overview

Endpoint	Method	Description
`/audio/speech`	POST	Text-to-speech — returns raw audio bytes
`/audio/transcriptions`	POST	Speech-to-text — returns a transcript

Both endpoints are compatible with the OpenAI SDK. Point your client at https://api.aicredits.in/v1 and your existing audio code works without changes.

Text-to-Speech

Send text and receive an audio file. The response is raw audio bytes — write them directly to a file or stream them to a media player.

from pathlib import Path
from openai import OpenAI

client = OpenAI(
    base_url="https://api.aicredits.in/v1",
    api_key="sk-your-key-here",
)

response = client.audio.speech.create(
    model="openai/tts-1",
    voice="alloy",
    input="Hello! Welcome to AICredits, your unified LLM gateway.",
)

Path("speech.mp3").write_bytes(response.content)

import OpenAI from "openai";
import fs from "fs";

const client = new OpenAI({
  baseURL: "https://api.aicredits.in/v1",
  apiKey: "sk-your-key-here",
});

const response = await client.audio.speech.create({
  model: "openai/tts-1",
  voice: "alloy",
  input: "Hello! Welcome to AICredits, your unified LLM gateway.",
});

const buffer = Buffer.from(await response.arrayBuffer());
fs.writeFileSync("speech.mp3", buffer);

curl https://api.aicredits.in/v1/audio/speech \
  -H "Authorization: Bearer sk-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/tts-1",
    "voice": "alloy",
    "input": "Hello! Welcome to AICredits, your unified LLM gateway."
  }' \
  --output speech.mp3

TTS Parameters

Parameter	Type	Description
`model`	string (required)	`openai/tts-1` for speed, `openai/tts-1-hd` for quality
`input`	string (required)	Text to synthesize. Maximum 4096 characters.
`voice`	string (required)	One of: `alloy`, `echo`, `fable`, `onyx`, `nova`, `shimmer`
`response_format`	string	Default: `mp3`. Also supports `opus`, `aac`, `flac`.

Voices

Voice	Character
`alloy`	Neutral and balanced — good all-purpose default
`echo`	Clear and crisp — well-suited for narration
`fable`	Expressive and warm — great for storytelling
`onyx`	Deep and authoritative — strong, confident tone
`nova`	Friendly and upbeat — energetic, conversational
`shimmer`	Soft and gentle — calm, soothing delivery

Transcription

Transcribe audio files to text using Whisper:

from openai import OpenAI

client = OpenAI(
    base_url="https://api.aicredits.in/v1",
    api_key="sk-your-key-here",
)

response = client.audio.transcriptions.create(
    model="openai/whisper-1",
    file=("audio.mp3", open("audio.mp3", "rb"), "audio/mp3"),
)

print(response.text)

AUDIO_B64=$(base64 -i audio.mp3)

curl https://api.aicredits.in/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-your-key-here" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"openai/whisper-1\",
    \"input_audio\": {
      \"data\": \"$AUDIO_B64\",
      \"format\": \"mp3\"
    }
  }"

Transcription Parameters

Parameter	Type	Description
`model`	string (required)	`openai/whisper-1` is currently supported
`input_audio.data`	string (required)	Base64-encoded audio content
`input_audio.format`	string (required)	`mp3`, `mp4`, `mpeg`, `mpga`, `m4a`, `wav`, `webm`
`language`	string	ISO-639-1 code (e.g. `en`, `hi`). Auto-detected if omitted.

Sarvam AI — Indian Languages

Sarvam AI is an India-first voice provider specialising in Indian languages. AICredits routes sarvam/bulbul-v2 and sarvam/bulbul-v3 (TTS) and sarvam/saarika-v2 (STT) directly to the Sarvam API — no extra setup beyond your AICredits key.

Sarvam voices are called speakers and use Indian names (e.g. anushka, rahul). Pass the speaker name in the voice field — or use an OpenAI alias (alloy, nova, etc.) and AICredits maps it to a Sarvam speaker automatically.

Sarvam TTS

Hindi TTS with Sarvam

curl https://api.aicredits.in/v1/audio/speech \
  -H "Authorization: Bearer sk-your-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sarvam/bulbul-v2",
    "input": "नमस्ते, आप कैसे हैं?",
    "voice": "anushka"
  }' \
  --output speech.wav

Sarvam STT

Hindi transcription with Sarvam

AUDIO_B64=$(base64 -i speech.wav)

curl https://api.aicredits.in/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-your-key-here" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"sarvam/saarika-v2\",
    \"input_audio\": {
      \"data\": \"$AUDIO_B64\",
      \"format\": \"wav\"
    },
    \"language\": \"hi-IN\"
  }"

OpenAI Voice Alias Mapping

OpenAI Alias	Sarvam Speaker
`alloy`	`anushka`
`echo`	`abhilash`
`fable`	`arya`
`onyx`	`karun`
`nova`	`priya`
`shimmer`	`shreya`

All available Sarvam speakers: anushka, abhilash, manisha, vidya, arya, karun, hitesh, aditya, ritu, priya, neha, rahul, pooja, rohan, simran, kavya, amit, dev, ishita, shreya, ratan, varun, and more.

Language Codes

Pass a BCP-47 language code for Sarvam models. Defaults to hi-IN if omitted:

Code	Language
`hi-IN`	Hindi
`en-IN`	English (India)
`bn-IN`	Bengali
`gu-IN`	Gujarati
`kn-IN`	Kannada
`ml-IN`	Malayalam
`mr-IN`	Marathi
`pa-IN`	Punjabi
`ta-IN`	Tamil
`te-IN`	Telugu

Models

Use the right model for the right endpoint

TTS models only work with /audio/speech. Transcription models only work with /audio/transcriptions. Mixing them returns an error.

Text-to-Speech — use with /audio/speech:

Model ID	Best For
`openai/tts-1`	Low latency, real-time synthesis
`openai/tts-1-hd`	Higher quality audio — podcasts, voiceovers
`sarvam/bulbul-v2`	Indian language TTS — 12 languages, 44 speakers

Transcription — use with /audio/transcriptions:

Model ID	Best For
`openai/whisper-1`	Multilingual speech recognition and translation
`sarvam/saarika-v2`	Indian language STT — optimised for Indian accents

Billing

Model	Unit	Rate (USD)
`openai/tts-1`	Per 1K characters	$0.015
`openai/tts-1-hd`	Per 1K characters	$0.030
`openai/whisper-1`	Per minute of audio	$0.006
`sarvam/bulbul-v2`	Per request	INR-native pricing
`sarvam/saarika-v2`	Per second of audio	INR-native pricing

Error Handling

Status	Cause
400	Missing required field (model, input, or voice)
400	Input exceeds 4096 characters
402	Insufficient wallet balance
429	Rate limit exceeded — audio requests count toward your RPM limit
502	All upstream providers failed — retry with exponential backoff

On this page