PlatterHubPlatterHub runs influencer campaigns for brands. Our voice answers its brands and creators: campaign questions, onboarding and support, spoken, not ticketed.
platterhub.aiQuick TTS speaks natural US English from a model we trained and an engine we wrote. No vendor sits in the middle taking a margin, so you pay a fraction of what voice costs you today. One API call and it is talking.
The bill for these words
Press play to start the meter
Write a line your product would say, pick a voice and listen. A sentence costs us a few thousandths of a cent, so go ahead.
What a million characters of speech costs at list price, taken from each company's own pricing page.
$2others average $31
$20others average $308
$200others average $3,075
List prices per million characters from each company's public pricing page, checked on 15 November 2026: OpenAI tts-1; Deepgram Aura-2, pay as you go; Cartesia Sonic, Scale plan overage; ElevenLabs Flash / Turbo and Multilingual v2. "Others average" is the mean of OpenAI tts-1, Deepgram Aura-2, Cartesia Sonic and ElevenLabs Flash. Plans, models and discounts change, so check their pages for today's price. All names belong to their owners.
Post your text to one endpoint and the audio starts coming back while the rest is still being made. Nothing to install and nothing new to learn.
curl -sN -X POST "$QUICK_TTS_URL/v1/tts" \
-H "x-api-key: $QUICK_TTS_KEY" \
-H "content-type: application/json" \
-d '{"text":"Your table for four is confirmed for eight tonight.",
"voice":"kiara","format":"wav"}' \
-o hello.wav
import os, requests
r = requests.post(
os.environ["QUICK_TTS_URL"] + "/v1/tts",
headers={"x-api-key": os.environ["QUICK_TTS_KEY"]},
json={"text": "Your table for four is confirmed for eight tonight.",
"voice": "kiara", "format": "wav"},
stream=True,
)
r.raise_for_status()
with open("hello.wav", "wb") as f:
for chunk in r.iter_content(chunk_size=None): # bytes arrive as they are made
f.write(chunk)
const res = await fetch(process.env.QUICK_TTS_URL + "/v1/tts", {
method: "POST",
headers: { "x-api-key": process.env.QUICK_TTS_KEY,
"content-type": "application/json" },
body: JSON.stringify({
text: "Your table for four is confirmed for eight tonight.",
voice: "kiara", format: "wav" }),
});
if (!res.ok) throw new Error(`${res.status} ${await res.text()}`);
// res.body is a stream: play or forward it as it arrives
Your endpoint and key arrive with your pilot.
to make a full minute of speech
live conversations on one four-core server
faster than real time, per server
Measured on one four-core cloud server on 3 November 2026, with paragraph-length text and no gaps in playback. We publish below what we measure, so there is headroom on every figure.
Tell us what you are building. You get an endpoint and a key, and you hear the difference on your own requests.
Your message goes straight to the two founders.
Our first pilot partner puts the voice in front of real brands and creators. The next places are open now.
Apply for a pilot
PlatterHubPlatterHub runs influencer campaigns for brands. Our voice answers its brands and creators: campaign questions, onboarding and support, spoken, not ticketed.
platterhub.aiAgentBox is a voice agent for your website and your phone line, built on the same voices and the same engine. It listens, gets the job done in your systems and hands over to a person when it should.
Anything missing? Ask in the pilot form and a founder will reply.
A text-to-speech API built on our own voice model and our own engine. You send text over REST or WebSocket and natural US English speech comes back while it is still being generated.
$2 per million characters, which works out at about 0.2¢ for a minute of speech. There are no seats, no setup fee and no charge per stream.
We trained the model and wrote the engine ourselves, and it runs on ordinary CPU servers. There is no model vendor or GPU bill in the middle, so the saving goes to you.
The first audio leaves our server 11 to 64 milliseconds after your text arrives, and a full minute of speech takes about ten seconds to make.
Eighteen US English voices at 24 kHz, with the pace set anywhere from 0.6× to 1.8× on each request. You can try them in the playground on this page.
Yes. Currency amounts, dates, phone numbers and acronyms are read the way a person would say them.
One four-core server carries 20 live streams. Capacity grows by adding servers, and a stream that has started is never cut short.
Yes. It runs on ordinary CPU servers, in our cloud or inside your own data centre, so your data can stay where it is.
Yes. We need about thirty minutes of clean studio audio, with the speaker's consent on record, and two to three weeks. It costs the same per character and runs at the same speed.
Apply for a pilot. We reply by email with an endpoint, a key and a short quickstart.