OpenAI Compatibility

murmr is designed as a drop-in replacement for the OpenAI TTS API. Migrate your existing code with minimal changes and unlock powerful new features.

Quick Migration

The fastest migration path uses VoiceDesign — describe any voice in natural language instead of choosing from preset names:

curl -X POST "https://api.murmr.dev/v1/voices/design" \
  -H "Authorization: Bearer YOUR_MURMR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello, world!",
    "voice_description": "A warm, professional female voice",
    "language": "English"
  }' --output output.wav

OpenAI Voice Equivalents

Instead of mapping OpenAI voices 1:1, describe what you want:

OpenAI alloy → "A neutral, clear voice for general use"
OpenAI echo → "A warm, resonant male voice"
OpenAI nova → "A bright, energetic female voice"
OpenAI shimmer → "A playful, upbeat voice"

Voice Strategy

Unlike OpenAI's fixed voices, murmr uses VoiceDesign — describe any voice in natural language and generate speech with it. Here's the recommended migration approach:

Step 1Create voices with VoiceDesign

Use the Playground or VoiceDesign API to describe the voice you want: "A warm, professional female voice, calm and clear"

Step 2Save voices you like

Found a voice you love? Save it via the Voices API to get a stable ID like voice_abc123

Step 3Use saved voices in production

Use your saved voice IDs with the batch or streaming endpoints for consistent, repeatable output.

Key Differences

SameEndpoint structure & auth

/v1/audio/speech with Bearer token in Authorization header — same as OpenAI.

BetterVoice flexibility

Instead of 6 fixed voices, create any voice with VoiceDesign. Describe exactly what you need.

BetterStreaming support

SSE streaming for low-latency playback (~450ms TTFC), or batch mode for complete audio files.

ExtraResponse format options

Batch endpoint supports response_format: mp3, opus, aac, flac, wav (default), pcm

ExtraWebSocket real-time

WebSocket API for voice agents and LLM integration. See Real-time docs.

SameSync batch

/v1/audio/speech returns 200 with binary audio by default — same as OpenAI. For async processing, pass webhook_url to get a 202 with a job ID. For low-latency streaming, use /v1/audio/speech/stream (~450ms to first audio).

DifferentLanguage parameter

murmr adds a language parameter using full names ("English", "French", "Japanese"). Defaults to "Auto" for automatic detection. See Language Support.

OpenAI SDK Compatibility

/v1/audio/speech now returns 200 with binary audio by default — the OpenAI SDK's client.audio.speech.create() works as a drop-in (with a saved voice ID as the voice parameter).

After: murmr with Saved Voice

Once you've saved a voice, use it with the streaming endpoint for real-time playback or the batch endpoint for bulk generation:

Streaming (recommended for real-time)

curl -X POST "https://api.murmr.dev/v1/audio/speech/stream" \
  -H "Authorization: Bearer YOUR_MURMR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello!",
    "voice": "voice_abc123",
    "language": "English"
  }'
# → SSE stream with base64 PCM chunks (~450ms to first audio)

Batch (for bulk generation)

# Returns 200 with binary audio
curl -X POST "https://api.murmr.dev/v1/audio/speech" \
  -H "Authorization: Bearer YOUR_MURMR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello!",
    "voice": "voice_abc123",
    "response_format": "mp3"
  }' --output hello.mp3

Both text and input field names are accepted for OpenAI compatibility.

Voice Mapping

Create a replacement voice for each OpenAI voice you use. This is a one-time setup — save the voice ID and use it in production:

// One-time setup: create and save your voices
const wav = await client.voices.design({
  input: 'This is a reference recording for the Nova replacement voice.',
  voice_description: 'A warm, friendly female voice, mid-20s, American',
});

const saved = await client.voices.save({
  name: 'Nova Replacement',
  description: 'A warm, friendly female voice, mid-20s, American',
  audio: wav,
  ref_text: 'This is a reference recording for the Nova replacement voice.',
});

console.log(`Nova replacement: ${saved.id}`);

Pricing

murmr offers competitive pricing compared to OpenAI TTS, especially at higher volumes. Check the Pricing page for details.