murmr
OpenAI-Compatible API

Stop settling for robotic TTS.
Design voices that sound human.

Describe the voice you want in plain English — "warm narrator, slight British accent, podcast-ready" — and start streaming in 97ms. No ML experience needed.

Free tier available. No credit card.

Ship in 5 minutes

Drop-in replacement for OpenAI TTS. Swap one line, keep your code.

voice-design.py
# Design a voice in natural language
voice = murmr.design(
  instruction="Warm female voice,
    soft Mexican accent, friendly tone",
  language="es"
)

# Generate speech
audio = murmr.speak(
  text="Hola, ¿cómo estás?",
  voice=voice
)|

Built for indie builders

VoiceDesign™

Skip the voice marketplace. Describe what you need in natural language — age, tone, accent, energy — and generate it instantly.

10 Real Languages

Spanish, Portuguese, German, French, Japanese, Korean, Chinese, Russian, Italian, English. Native speakers, not accented English.

ESPTDEFRJAKOZHRUITEN

97ms First Byte

Real-time streaming for voice agents, games, and interactive apps. Fast enough for conversation.

97ms

Self-Host Option

Run on your own GPU. Your audio, your data, your infrastructure. Perfect for sensitive use cases.

What people are building

Voice agents & chatbotsIndie game dialoguePodcast automationAudiobook narrationE-learning coursesAccessibility tools
Live Demo

Try it now

Describe a voice and hear it speak. No signup required.

Try it now

Describe a voice and hear it speak. No signup required.

74/500

Pricing that makes sense

Start free. Scale when you need to.

Free

$0

50K characters/month

  • 3 voice designs
  • All 10 languages
  • Community Discord
Best Value

Starter

$9/mo

200K characters/month

  • 10 voice designs
  • Commercial license
  • Email support

Pro

$29/mo

1M characters/month

  • Unlimited designs
  • Priority support
  • 99.9% uptime SLA

Be first in line

Join developers shipping voice features without the complexity.

Frequently asked questions

What is VoiceDesign?
VoiceDesign lets you create custom voices by describing them in plain English. Instead of choosing from a preset list or uploading audio samples, just describe what you want: 'warm male narrator, slight British accent, calm energy' — and we generate it instantly.
How is this different from voice cloning?
Voice cloning copies an existing voice from audio samples. VoiceDesign creates entirely new voices from text descriptions. No recordings needed, no consent issues, no licensing complexity.
What languages do you support?
10 languages with native speaker quality: English, Spanish, Portuguese, German, French, Italian, Chinese, Japanese, Korean, and Russian. Each voice sounds natural in its target language — not accented English.
Is there a free tier?
Yes. The free plan includes 50,000 characters per month, 3 voice designs, and access to all 10 languages. No credit card required.
Can I self-host murmr?
Yes. Our API is fully open-source and self-hostable. Run it on your own GPU for complete data privacy and control. Perfect for sensitive applications or air-gapped environments.
Is the API compatible with OpenAI?
Yes. murmr is a drop-in replacement for OpenAI's TTS API. Same endpoint structure, same request format. Swap one line of code to switch providers.