Lowest cost
audio generation API.
One endpoint for speech, music, and sound effects. Open source on Runware hardware, closed source at market rates. Pay per generation, no subscriptions.
Every audio model, one API.
Speech, music, and sound effects on a single endpoint, from every leading provider. Pay per generation, with no subscriptions to commit to.
Start with the outcome you need.
One endpoint covers speech, music, and sound effects. Pick what you're making, and start from the model built for it.
Speech / TTS
Turn text into natural, expressive speech for narration, voice agents, dubbing, and in-app voice, across dozens of languages.
Flight log voiceover
- Expressive narration
fishaudio:s2.1@pro / google:[email protected]Fine control over emotion and pacing, with multi-speaker dialogue for narration, podcasts, and audiobooks.
- Low-latency / interactive voice
inworld:tts@2Tuned for conversational use in assistants and agents. Inworld Realtime TTS-2 holds one voice identity across 100+ languages; TTS-1.5 Mini is built for ultra-low-latency interactive apps.
- Multilingual
google:[email protected] / fishaudio:s2.1@proWide language coverage with automatic language detection, for products shipping across markets.
Music
Generate full tracks from a prompt, vocal songs or instrumentals, with control over structure, genre, and style.
Crystalline electronic ambient track
- Full songs (vocal + instrumental)
minimax:[email protected]Full songs or instrumentals from natural-language prompts. Follows BPM and key, with section-level structure control.
- Song cover / restyle
minimax:music@coverReimagine an existing track in a new style. Keeps the vocal melody while changing timbre, instrumentation, and genre from a prompt.
- Open-source music at length
runware:[email protected]Open-source generation for longer compositions, with cover, remix, and lyric editing.
Sound effects
Generate ambient sound, effects, and textures from a text description.
Cooking eggs in a frying pan
- Cinematic SFX
mirelo:1@1Describe a sound and get a usable effect for UI cues, game audio, and atmosphere.
- Versatile text-to-audio
bytedance:[email protected]Generate expressive voice, music and SFX sounds in one call.
Voice cloning & design
Build a custom voice from a short sample or a written description.
A cheerful, young American woman.
- Clone from a sample
alibaba:[email protected] / alibaba:[email protected]Clone a voice from a ~3-second sample, in 10+ languages.
- Design from a description
alibaba:[email protected]Build a voice from a written description of emotion, tone, and prosody.
One account. Top providers.
Every leading audio provider on one endpoint, with the same auth and the same bill.Switching model is a string change.
MiniMax
Music and speech in one house.
MiniMax Music 2.6: full songs & instrumentals
minimax:[email protected]MiniMax Music Cover: song-to-song restyle
minimax:music@coverMiniMax Speech 2.8: broadcast-ready TTS
minimax:[email protected]Inworld
Low-latency, conversational speech.
Inworld Realtime TTS-2: conversational, 100+ languages
inworld:tts@2Inworld TTS-1.5 Max: high-fidelity voice
inworld:[email protected]Inworld TTS-1.5 Mini: ultra-low latency
inworld:[email protected]Alibaba (Qwen-TTS)
Human-like speech, cloning, and voice design.
Qwen3-TTS 1.7B Base: clone from a 3s sample
alibaba:[email protected]Qwen3-TTS 1.7B CustomVoice: preset timbres, style control
alibaba:[email protected]Qwen3-TTS 1.7B VoiceDesign: voice from a description
alibaba:[email protected]Fish Audio
Expressive, low-latency text to speech.
Fish Audio S2.1 Pro: expressive, 80+ languages
fishaudio:s2.1@proGoogle (Gemini TTS)
Expressive speech with fine delivery control.
Gemini 3.1 Flash TTS: audio tags, 70+ languages
google:[email protected]
xAI
Natural spoken audio in one call.
xAI Text-to-Speech: five voices, 20+ languages
xai:tts@0ByteDance
Speech and audio generation.
Seed Audio 1.0: expressive TTS, voice conversion, speech editing
bytedance:[email protected]Any workflow, one request.
A single endpoint covers every audio workflow. Pass the relevant inputs and Runware routes the request to the right one. No need to switch endpoints to switch task type.
Pass speech.text
Pass the text to speak and the model returns natural speech. The simplest workflow.
{
"taskType": "audioInference",
"taskUUID": "...",
"model": "fishaudio:s2.1@pro",
"speech": {
"text": "Welcome to Runware. Every audio model, one API."
}
}Built for production. Not just demos.
One integration covers speech, music, and sound effects. These are the applications teams build on it.
FAQ
Get an API key.
One endpoint for speech, music, and sound effects.Pay per generation, no subscriptions.