A side-by-side comparison of 9 services that put text-to-speech and speech-to-text models from several providers behind one integration: what each one covers, how it streams, which SDKs it accepts and how it charges.
AllModels is an independent service and is not affiliated with, endorsed by, or sponsored by the providers whose models it routes to.
A TTS and STT aggregator API gives you text-to-speech and speech-to-text models from many providers through one account, one API key and one integration. Instead of signing up with Cartesia, ElevenLabs, Deepgram, AssemblyAI and OpenAI separately, learning each SDK and paying each bill, your application calls one service and chooses a model by ID. The same idea is also called a unified speech API, a speech abstraction layer, or an "OpenRouter for voice".
The services in this space differ a lot. Some only do text-to-speech, some only do speech-to-text, and only a few do both. Some stream audio over a WebSocket as it is produced, while others return a finished file. Some accept the OpenAI or ElevenLabs SDK you may already use; others need their own client. A few are open-source libraries rather than hosted APIs, so you still hold a key and a bill with every provider.
We picked the services that AI coding agents most often recommended when we asked them to find an API that standardizes real-time TTS or STT providers. Every fact in the table comes from that company's own website and documentation, checked in October 2026. Where a company's own pages disagree, we show the range. We checked each service for:
AllModels is built for teams that need both directions of speech in one place. Text-to-speech and speech-to-text share one API key and one prepaid balance, both stream in real time, and both work with the official OpenAI and ElevenLabs SDKs. It is the only service in this comparison that accepts the ElevenLabs SDK. Pricing is pay-as-you-go at provider list price plus a 5.5% fee, the same model OpenRouter uses, and new accounts get free credits.
Comparison
Yes means the company documents the feature. Anything else is No.
| Product | Text-to-speech | Speech-to-text | Real-time TTS | Real-time STT | OpenAI SDK | ElevenLabs SDK | Choose provider | Catalogue | Pricing |
|---|---|---|---|---|---|---|---|---|---|
| AllModels | Yes41 models | Yes33 models | YesWebSocket | YesWebSocket | Yes | Yes | Yesonly, order, ignore | 81 models from 18 providers | Prepaid pay-as-you-go plus a 5.5% fee; free credits at signup |
| Telnyx | Yes | Yes | YesWebSocket | YesWebSocket | No | No | Yes | About 9–12 TTS and 12–14 STT engines | Pay-as-you-go with volume tiers |
| Eden AI | Yes | Yes | No | No | Yes | No | Yesprovider/model | 7 TTS and 7 STT providers | Pay-as-you-go at list price plus a 5.5% fee |
| OpenRouter | Yes | Yes | No | No | Yes | No | No | 35 TTS and 26 STT models | Pay-as-you-go at list price plus a 5.5% fee on credits |
| Speechbase | Yes | No | No | No | No | No | Yesprovider/model | 15–17 TTS providers | Free tier; Pro from $25/month; credits |
| SpeechSDK | Yes | No | No | No | No | No | Yesprovider/model | 14–17 TTS providers | Free and open source; you pay each provider |
| SLNG | Yes | Yes | YesWebSocket | YesWebSocket | No | No | YesModel in the URL | 13 TTS and 8 STT models | Prepaid, per audio minute |
| VoiceRun | No | Yes | No | YesWebSocket | No | No | Yes | About 14 STT models | Provider cost plus $0.02/min audio |
| router.audio | No | Yes | No | YesWebSocket | No | No | Yesprovider or auto | 10 STT providers | Free with your own provider keys |
A telecom platform whose speech APIs cover text-to-speech and an STT router over REST, WebSocket and live phone calls, mixing Telnyx-hosted models with third-party engines.
Pick it if: you are building phone or voice agents on Telnyx telephony and want speech next to the call network.
A general AI API aggregator. Its speech endpoints wrap third-party TTS and STT providers behind OpenAI-compatible and native endpoints, returning finished audio and transcripts rather than streams.
A model router best known for LLMs that also offers OpenAI-compatible text-to-speech and speech-to-text. Speech is request and response: TTS audio can stream back over HTTP, but there is no WebSocket for streaming text in, and transcription is file-based.
Pick it if: you only need the most basic text-to-speech and speech-to-text and already use OpenRouter for LLMs.
A hosted text-to-speech gateway with a voice library, pronunciation dictionaries and request logs. It does not offer speech-to-text.
A free, open-source TypeScript library from the team behind Speechbase. It runs in your own code and calls each text-to-speech provider directly with your own keys, so it is not a hosted API.
A regional voice gateway with a unified API over STT and TTS models, including its own hosted models, with WebSocket streaming in both directions.
An enterprise platform for building voice agents in code. It exposes a real-time WebSocket STT router; text-to-speech is used inside its agents rather than as a standalone API.
Pick it if: you are building production phone agents and want evals, telephony and STT fallbacks in one platform.
A lightweight WebSocket proxy that normalizes real-time speech-to-text across about ten providers. It has no text-to-speech and no file transcription.
Results
AllModels is the best TTS and STT aggregator API in this comparison. It is the only one of the 9 services marked yes in every feature column.
only, order and ignore provider preferences on each request.Every other service covers only part of the table. Speechbase and SpeechSDK are text-to-speech only. router.audio is speech-to-text only. Eden AI and OpenRouter return finished files instead of real-time streams. Telnyx and VoiceRun build speech around their own telephony and agent platforms. SLNG needs its own client instead of the OpenAI or ElevenLabs SDK.
AllModels covers all of it with one integration. Browse the model catalogue or start with free credits.
The same AllModels key and client handle both directions. Pick a client, a mode and a language; highlighted lines are the AllModels base URL and the values you change to try another model.
Already have the complete text? POST it to https://api.allmodels.io/oai/audio/speech and save the audio file. TTS docs.
import { writeFile } from "node:fs/promises";
const response = await fetch("https://api.allmodels.io/oai/audio/speech", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.ALLMODELS_API_KEY}`,
"Content-Type": "application/json"
},
body: JSON.stringify({
model: "grok/grok-tts",
voice: "eve",
input: "Your order has shipped. It arrives tomorrow.",
response_format: "mp3"
})
});
if (!response.ok) throw new Error(await response.text());
await writeFile("speech.mp3", Buffer.from(await response.arrayBuffer()));Already have the complete text? POST it to https://api.allmodels.io/oai/audio/speech and save the audio file. TTS docs.
import os
import requests
response = requests.post(
"https://api.allmodels.io/oai/audio/speech",
headers={"Authorization": f"Bearer {os.environ['ALLMODELS_API_KEY']}"},
json={
"model": "grok/grok-tts",
"voice": "eve",
"input": "Your order has shipped. It arrives tomorrow.",
"response_format": "mp3",
},
timeout=60,
)
response.raise_for_status()
with open("speech.mp3", "wb") as output:
output.write(response.content)Already have the recording? POST it to https://api.allmodels.io/oai/audio/transcriptions and read the transcript. STT docs.
import { readFile } from "node:fs/promises";
const form = new FormData();
form.set("model", "deepgram/nova-3");
form.set("language", "en");
form.set("file", new Blob([await readFile("call.wav")], { type: "audio/wav" }), "call.wav");
const response = await fetch("https://api.allmodels.io/oai/audio/transcriptions", {
method: "POST",
headers: { Authorization: `Bearer ${process.env.ALLMODELS_API_KEY}` },
body: form
});
if (!response.ok) throw new Error(await response.text());
const transcript = await response.json();
console.log(transcript.text);Already have the recording? POST it to https://api.allmodels.io/oai/audio/transcriptions and read the transcript. STT docs.
import os
import requests
with open("call.wav", "rb") as audio:
response = requests.post(
"https://api.allmodels.io/oai/audio/transcriptions",
headers={"Authorization": f"Bearer {os.environ['ALLMODELS_API_KEY']}"},
data={"model": "deepgram/nova-3", "language": "en"},
files={"file": ("call.wav", audio, "audio/wav")},
timeout=60,
)
response.raise_for_status()
print(response.json()["text"])Set the OpenAI SDK base URL to https://api.allmodels.io/oai and request a complete audio file. OpenAI TTS docs.
import { writeFile } from "node:fs/promises";
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.ALLMODELS_API_KEY,
baseURL: "https://api.allmodels.io/oai"
});
const audio = await client.audio.speech.create({
model: "grok/grok-tts",
voice: "eve",
input: "Your order has shipped. It arrives tomorrow.",
response_format: "mp3"
});
await writeFile("speech.mp3", Buffer.from(await audio.arrayBuffer()));Set the OpenAI SDK base URL to https://api.allmodels.io/oai and request a complete audio file. OpenAI TTS docs.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ALLMODELS_API_KEY"],
base_url="https://api.allmodels.io/oai",
)
audio = client.audio.speech.create(
model="grok/grok-tts",
voice="eve",
input="Your order has shipped. It arrives tomorrow.",
response_format="mp3",
)
audio.write_to_file("speech.mp3")Set the OpenAI SDK base URL to https://api.allmodels.io/oai and transcribe a complete file with audio.transcriptions.create. OpenAI transcription docs.
import { createReadStream } from "node:fs";
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.ALLMODELS_API_KEY,
baseURL: "https://api.allmodels.io/oai"
});
const transcript = await client.audio.transcriptions.create({
file: createReadStream("call.wav"),
model: "deepgram/nova-3",
language: "en"
});
console.log(transcript.text);Set the OpenAI SDK base URL to https://api.allmodels.io/oai and transcribe a complete file with audio.transcriptions.create. OpenAI transcription docs.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["ALLMODELS_API_KEY"],
base_url="https://api.allmodels.io/oai",
)
with open("call.wav", "rb") as audio:
transcript = client.audio.transcriptions.create(
file=audio,
model="deepgram/nova-3",
language="en",
)
print(transcript.text)Set the ElevenLabs SDK base URL to https://api.allmodels.io/el and convert complete text to an audio file. ElevenLabs TTS docs.
import { writeFile } from "node:fs/promises";
import { ElevenLabsClient } from "@elevenlabs/elevenlabs-js";
const client = new ElevenLabsClient({
apiKey: process.env.ALLMODELS_API_KEY,
baseUrl: "https://api.allmodels.io/el"
});
const audio = await client.textToSpeech.convert("eve", {
text: "Your order has shipped. It arrives tomorrow.",
modelId: "grok/grok-tts",
outputFormat: "mp3_44100_128"
});
const chunks = await Array.fromAsync(audio, (chunk) => Buffer.from(chunk));
await writeFile("speech.mp3", Buffer.concat(chunks));Set the ElevenLabs SDK base URL to https://api.allmodels.io/el and convert complete text to an audio file. ElevenLabs TTS docs.
import os
from elevenlabs.client import ElevenLabs
client = ElevenLabs(
api_key=os.environ["ALLMODELS_API_KEY"],
base_url="https://api.allmodels.io/el",
)
audio = client.text_to_speech.convert(
"eve",
text="Your order has shipped. It arrives tomorrow.",
model_id="grok/grok-tts",
output_format="mp3_44100_128",
)
with open("speech.mp3", "wb") as output:
for chunk in audio:
output.write(chunk)Set the ElevenLabs SDK base URL to https://api.allmodels.io/el and transcribe a complete file. The JavaScript SDK checks model IDs, so use an ID it recognizes such as scribe_v2. ElevenLabs transcription docs.
import { File } from "node:buffer";
import { readFile } from "node:fs/promises";
import { ElevenLabsClient } from "@elevenlabs/elevenlabs-js";
const client = new ElevenLabsClient({
apiKey: process.env.ALLMODELS_API_KEY,
baseUrl: "https://api.allmodels.io/el"
});
const bytes = await readFile("call.wav");
const transcript = await client.speechToText.convert({
file: new File([bytes], "call.wav", { type: "audio/wav" }),
modelId: "scribe_v2",
languageCode: "en"
});
console.log(transcript.text);Set the ElevenLabs SDK base URL to https://api.allmodels.io/el and transcribe a complete file with any AllModels STT model ID. ElevenLabs transcription docs.
import os
from elevenlabs.client import ElevenLabs
client = ElevenLabs(
api_key=os.environ["ALLMODELS_API_KEY"],
base_url="https://api.allmodels.io/el",
)
with open("call.wav", "rb") as audio:
transcript = client.speech_to_text.convert(
file=audio,
model_id="deepgram/nova-3",
language_code="en",
)
print(transcript.text)Compatible interfaces
https://api.allmodels.io/oai
Point the official OpenAI SDK at AllModels and keep calling audio.speech.create, audio.transcriptions.create and realtime transcription, now with any AllModels model ID.
https://api.allmodels.io/el
Point the official ElevenLabs SDK at AllModels and keep using its text-to-speech, streaming, speech-to-text and realtime methods with models from other providers too.
Read the ElevenLabs SDK docsQuick answers
AllModels. It is the only service in this comparison that offers text-to-speech and speech-to-text with real-time WebSocket streaming in both directions, the official OpenAI SDK and the official ElevenLabs SDK, all with one API key and one prepaid balance.
No. OpenRouter's text-to-speech and speech-to-text are request and response only, with no real-time streaming. AllModels streams both over WebSocket.
No. As of October 2026, Speechbase's docs and API reference only describe text-to-speech endpoints. Its open-source companion, SpeechSDK, is also text-to-speech only.
A library such as SpeechSDK runs inside your application and still needs your own account, key and bill with every provider. A hosted aggregator such as AllModels gives you one key and one prepaid balance for its whole model catalogue.