Sarvam AI 语音转文本 API：支持23种语言、说话人分离与流式传输的语音识别解决方案

speech-to-text by sarvamai/skills

137 周安装量

57 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/sarvamai/skills --skill speech-to-text

AI/机器学习音频处理 API

🇨🇳中文介绍

语音转文本 — Saaras

[!重要] 认证：api-subscription-key 请求头 — 而非 Authorization: Bearer。基础 URL：https://api.sarvam.ai/v1

模型

saaras:v3 — 支持 23 种语言，5 种输出模式（transcribe、translate、verbatim、translit、codemix），自动语言检测。

快速开始 (Python)

广告位招租

在这里展示您的产品或服务

触达数万 AI 开发者，精准高效

联系我们

快速开始 (JavaScript/TypeScript)

import { SarvamAIClient } from "sarvamai";
import * as fs from "fs";

const client = new SarvamAIClient({ apiSubscriptionKey: "YOUR_SARVAM_API_KEY" });

const response = await client.speechToText.transcribe({
    file: fs.createReadStream("audio.wav"),
    model: "saaras:v3",
    mode: "transcribe"
});
console.log(response.transcript);

批量 API (长音频 + 说话人分离)

job = client.speech_to_text_job.create_job(
    model="saaras:v3",
    mode="transcribe",
    language_code="hi-IN",
    with_diarization=True,
    num_speakers=2
)
job.upload_files(file_paths=["meeting.mp3"])
job.start()
job.wait_until_complete()
job.download_outputs(output_dir="./output")

支持长达 1 小时的音频，最多 8 位说话人，所有 5 种输出模式。

WebSocket 流式传输

import asyncio, base64
from sarvamai import AsyncSarvamAI

async def stream_audio():
    client = AsyncSarvamAI()
    async with client.speech_to_text_streaming.connect(
        model="saaras:v3",
        high_vad_sensitivity=True,
        flush_signal=True
    ) as ws:
        with open("audio.wav", "rb") as f:
            audio_base64 = base64.b64encode(f.read()).decode("utf-8")
        await ws.transcribe(audio=audio_base64, encoding="audio/wav", sample_rate=16000)
        await ws.flush()
        response = await ws.recv()
        print(response)

asyncio.run(stream_audio())

支持长达 8 小时的会话。电话音频请使用 sample_rate=8000。

注意事项	详情
REST：30 秒限制	音频 >30 秒会失败。处理长文件请使用批量 API 或 WebSocket。
JS 方法名	`client.speechToText.transcribe({...})` — 使用驼峰命名法，而非 `speech_to_text`。文件通过 `fs.createReadStream()` 传递。
WebSocket 编解码器	仅支持 `wav`、`pcm_s16le`、`pcm_l16`、`pcm_raw`。流式传输不支持 MP3/AAC/OGG。
WebSocket 音频	必须进行 base64 编码。电话音频请使用 `sample_rate=8000`。
刷新信号	`flush_signal=True` 结合 `await ws.flush()` 强制立即进行转录边界划分。
短音频检测	对于小于 3 秒的音频，请显式设置 `language_code` — 自动检测需要更多信号。

获取流式传输协议、批量 API SDK 示例和编解码器详细信息，请访问：

🇺🇸English

Speech-to-Text — Saaras

[!IMPORTANT] Auth: api-subscription-key header — NOT Authorization: Bearer. Base URL: https://api.sarvam.ai/v1

Model

saaras:v3 — 23 languages, 5 output modes (transcribe, translate, verbatim, translit, codemix), auto language detection.

Quick Start (Python)

from sarvamai import SarvamAI
client = SarvamAI()

response = client.speech_to_text.transcribe(
    file=open("audio.wav", "rb"),
    model="saaras:v3",
    mode="transcribe"
)
print(response.transcript)

Quick Start (JavaScript/TypeScript)

import { SarvamAIClient } from "sarvamai";
import * as fs from "fs";

const client = new SarvamAIClient({ apiSubscriptionKey: "YOUR_SARVAM_API_KEY" });

const response = await client.speechToText.transcribe({
    file: fs.createReadStream("audio.wav"),
    model: "saaras:v3",
    mode: "transcribe"
});
console.log(response.transcript);

Batch API (Long Audio + Diarization)

job = client.speech_to_text_job.create_job(
    model="saaras:v3",
    mode="transcribe",
    language_code="hi-IN",
    with_diarization=True,
    num_speakers=2
)
job.upload_files(file_paths=["meeting.mp3"])
job.start()
job.wait_until_complete()
job.download_outputs(output_dir="./output")

Supports audio up to 1 hour, up to 8 speakers, all 5 output modes.

WebSocket Streaming

import asyncio, base64
from sarvamai import AsyncSarvamAI

async def stream_audio():
    client = AsyncSarvamAI()
    async with client.speech_to_text_streaming.connect(
        model="saaras:v3",
        high_vad_sensitivity=True,
        flush_signal=True
    ) as ws:
        with open("audio.wav", "rb") as f:
            audio_base64 = base64.b64encode(f.read()).decode("utf-8")
        await ws.transcribe(audio=audio_base64, encoding="audio/wav", sample_rate=16000)
        await ws.flush()
        response = await ws.recv()
        print(response)

asyncio.run(stream_audio())

Supports sessions up to 8 hours. Use sample_rate=8000 for telephony audio.

Gotchas

Gotcha	Detail
REST: 30s limit	Audio >30s fails. Use Batch API or WebSocket for longer files.
JS method name	`client.speechToText.transcribe({...})` — camelCase, NOT `speech_to_text`. File via `fs.createReadStream()`.
WebSocket codecs	Only `wav`, `pcm_s16le`, `pcm_l16`, `pcm_raw`. MP3/AAC/OGG NOT supported for streaming.

Full Docs

Fetch streaming protocol, batch API SDK examples, and codec details from:

https://docs.sarvam.ai/llms.txt — comprehensive docs index
STT Overview
Streaming API
Batch API + Diarization
Rate Limits

Weekly Installs

Repository

sarvamai/skills

GitHub Stars

First Seen

Feb 12, 2026

Security Audits

Gen Agent Trust HubPass SocketPass SnykPass

Installed on

gemini-cli97

kimi-cli97

codex97

amp97

opencode97

github-copilot97

超能力技能使用指南：AI助手技能调用优先级与工作流程详解

52,100 周安装