CosyVoice

cosyvoice2-0.5b

CosyVoice2-0.5B

Low-latency speech synthesis for narration, assistants, support voice, and content dubbing.

Public model detailSpeech synthesis stack

Params

0.5B

Context

Voice synthesis

Max Output

Audio

License

Apache 2.0

TTFT

N/A

No 5m benchmark sample

5m RPM

N/A

Why pick it

  • About one-third of SiliconFlow audio pricing
  • Good default TTS route

Pricing

TierStandardCachedSiliconFlowSavings
Realtime$5.00 / 1M bytesN/A$15.0067%
BatchSame as standardN/A$15.0067%

Production pricing proof

How this route settles on a real request

When the model executes live, response headers can expose X-BatchIn-Provider, X-BatchIn-Route-Reason, X-BatchIn-Effective-Cost-Cents, and X-BatchIn-Uncached-Cost-Cents.
The cached price here means prompt-cache discount on input tokens. Durable response-cache hits are proven separately through X-BatchIn-Response-Cache-Mode and request lookup.
Streaming calls start with X-Request-Id, then resolve final cost, cache mode, and route truth through lookup after completion.
Open request lookupOpen trust hub

Route source of truth

See pricing, request proof, and the upgrade path on one page

Standard, prompt-cache, batch, and SiliconFlow comparison stay visible without leaving the route.
Real requests return X-Request-Id, and buffered calls can expose route reason, billed cost, and uncached cost directly.
BatchIn supports Playground validation first, then batch, white-label, or dedicated capacity conversations.
Talk to the teamOpen pricing

Quick start

OpenAI-compatible surface. Swap the base URL and ship.

Try in PlaygroundOpen pricingTalk to the team
Python
from openai import OpenAI

client = OpenAI(
    base_url="https://api.luminapath.tech/v1",
    api_key="BATCHIN_API_KEY"
)

resp = client.chat.completions.create(
    model="cosyvoice2-0.5b",
    messages=[{"role": "user", "content": "Summarize why this model is a fit for my workload."}]
)

print(resp.choices[0].message.content)
JavaScript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.luminapath.tech/v1",
  apiKey: process.env.BATCHIN_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "cosyvoice2-0.5b",
  messages: [{ role: "user", content: "Summarize why this model is a fit for my workload." }],
});

console.log(resp.choices[0]?.message?.content);
cURL
curl https://api.luminapath.tech/v1/chat/completions \
  -H "Authorization: Bearer $BATCHIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "cosyvoice2-0.5b",
    "messages": [{"role":"user","content":"Summarize why this model is a fit for my workload."}]
  }'

Specs

Architecture

Speech synthesis stack

Vendor group

Audio

Context window

Voice synthesis

Max output

Audio

Best for

audio
tts

Related models

Back to model center

Fish Audio

fish-speech-1-5

Fish-Speech-1.5

Open speech route for expressive synthesis, cloned voice styles, and content narration.

View detail

IndexTTS

indextts-2

IndexTTS-2

Fast speech stack for product voice output, IVR systems, and developer voice UX.

View detail

Alibaba

qwen3-tts

Qwen3-TTS

Alibaba’s lightweight TTS route for voice UX and product narration tasks.

View detail

Z.ai

glm-5.1

GLM-5.1

Open-source coding flagship built for long-horizon autonomous engineering and deep reasoning.

View detail
AI Assistant