EcoHash Inference
Text-to-Speech Studio
Synthesize speech with Kokoro-82M or Qwen3-TTS and see the measured latency, real-time factor and exact cost of each request. Both models run on EcoHash's own GPUs behind a single OpenAI-compatible endpoint.
Runs on the EcoHash demo key by default, with a small free allowance per visitor per day. Paste your own key to lift the limit — it is held for this browser session only, never stored, never logged, and never shown in the generated code below.
Kokoro-82M is an 82M-parameter model built for latency; Qwen3-TTS trades some speed for prosody and broader language coverage. Up to 800 characters per request in this demo.
- Kokoro-82M
- $1.00 / 1M tokens (≈ $0.006 per audio minute) · hexgrad/Kokoro-82M
- Qwen3-TTS
- $2.00 / 1M tokens (≈ $0.012 per audio minute) · Qwen/Qwen3-TTS-12Hz-1.7B-Base
Every model in this Space runs on EcoHash's own
NVIDIA RTX PRO 6000 Blackwell
fleet — no third-party resale — and is reachable through one OpenAI-compatible
endpoint at https://api.ecohash.com/v1. GPU instances from $1.89 per GPU-hour, billed per second.
Model catalog · Pricing · Docs · More demos · Create an account ($1 free credit)