Speech and audio

Turn text into speech and audio into text.

Generate speech with Kokoro 82M or Qwen3 TTS and transcribe audio with Whisper Large v3 through one API, priced per character or per minute.

Kokoro 82M$0.00104 per 1K characters

What you can build

Three jobs people run here. Each on a product you can start today.

Inference

Kokoro 82M for low-cost voices, Qwen3 TTS for multilingual speech, Orpheus 3B for expressive delivery. One OpenAI-compatible speech endpoint.

Inference

Whisper Large v3 and V3 Turbo transcribe meetings, podcasts and calls in many languages. Voxtral Small also answers questions about the audio.

Instance

Fine-tune a voice or run a speech model of your own on an L4 or L40S instance, with the weights on a volume.

Examples

Claude Code or Codex sends the request for you. Any OpenAI-compatible client works with one base URL.

Hand this to your agent

Add a Runcrate model to this project: run curl -fsSL"https://runcrate.ai/agents.md?intent=build"and follow it.

Your agent asks for an API key. Create a key

Works inClaude CodeCodexOpenCodeClineRoo CodeKilo CodeCursor

example

Models and tools

Name	Role	Runs on
Kokoro 82M	Low-cost text to speech	Inference API
Qwen3 TTS	Multilingual text to speech	Inference API
Orpheus 3B	Expressive text to speech	Inference API
Whisper Large v3	Multilingual transcription	Inference API
Whisper Large v3 Turbo	Faster transcription	Inference API
Voxtral Small	Transcription and audio understanding	Inference API

How it works

01
Choose your task
Text to speech for voices, speech to text for transcripts. Pick the model that fits your languages, quality and budget.
02
Call the API
Send text to the speech endpoint and get audio back, or send an audio file to the transcription endpoint and get text.
03
Build the pipeline
Chain speech models with a language model for voice agents, dubbing or call summaries. Speech is priced per character, transcription per minute.

Pricing

Name

As billed