Switch from Amazon Polly
Change the endpoint and the key; keep your code. Works with Polly, OpenAI and ElevenLabs clients for plain text to speech.
Amazon Polly (Python, boto3)
Change the client, keep the calls:
import boto3
polly = boto3.client("polly", endpoint_url="https://speakvora.com/api",
aws_access_key_id="YOUR_SPEAKVORA_KEY", aws_secret_access_key="unused", region_name="us-east-1")
r = polly.synthesize_speech(Text="Your order is ready.", VoiceId="Joanna", OutputFormat="mp3")
open("out.mp3", "wb").write(r["AudioStream"].read())Polly voice names map to the closest Speakvora catalog voice of the same gender and accent (Joanna to af_heart, Matthew to am_michael, Amy to bf_emma, Brian to bm_george). Speakvora voice ids also work.
OpenAI
from openai import OpenAI
client = OpenAI(base_url="https://speakvora.com/api/v1", api_key="YOUR_SPEAKVORA_KEY")
client.audio.speech.create(model="tts-1", voice="alloy", input="Your order is ready.")ElevenLabs
curl https://speakvora.com/api/v1/text-to-speech/af_heart -H "xi-api-key: YOUR_SPEAKVORA_KEY" \
-H "content-type: application/json" -d '{"text":"Your order is ready."}'What is different
- Audio comes back as mp3 or AAC (the content type says which).
- SSML, speech marks, lexicons and other formats return an error instead of being ignored.
- Requests are billed by the character at the normal Speakvora price, with the same spending caps.
Speed: text to speech latency. Prices: Amazon Polly pricing compared.
Questions
Can I replace Amazon Polly with Speakvora without rewriting my code?
For plain text to mp3 speech, yes. Point your existing Polly client at Speakvora's endpoint and use a Speakvora API key as the access key id. SynthesizeSpeech and DescribeVoices work. SSML, speech marks, lexicons and output formats other than mp3 or AAC are not supported and return an error.
Does it work with OpenAI and ElevenLabs clients too?
Yes for plain text to speech. The OpenAI client works with base URL https://speakvora.com/api/v1 and the ElevenLabs client with https://speakvora.com/api, each using your Speakvora key.
How much does it cost compared with Amazon Polly?
Speakvora catalog voices are $3 per million characters. Amazon Polly lists Standard $4, Neural $16, Generative $30, Long-Form $100 per million characters on its pricing page read 2026-10-08. Cost only; voices and languages differ.
How do I get an API key?
Sign in to your Speakvora account, open Voice API and choose Create key. New accounts get 100,000 characters free.
What do I lose by switching?
Polly's SSML control, speech marks, lexicons, more than 60 voices in more languages, other output formats and the AWS integrations. Speakvora is smaller and has no SLA.