Text To Speech error [Greg Kamradt's example]

[jonathan885]

Hi,
I just started with Deepgram and I'm trying to follow Greg Kamradt's example

Basically, for the text to speech mode, it does the following:

MODEL_NAME = "alpha-stella-en-v2"  # Example model name, change as needed
text = """The returns for performance are superlinear."""

DEEPGRAM_URL = f"https://api.beta.deepgram.com/v1/speak?model={MODEL_NAME}&performance=some&encoding=linear16&sample_rate=24000"

headers = {
        "Authorization": f"Token {config['DEEPGRAM_KEY']}",
        "Content-Type": "application/json"
    }
    
  payload = {
      "text": text,
      "voice": MODEL_NAME
  }

then loop over the chunks of the request

with requests.post(DEEPGRAM_URL, stream=True, headers=headers, json=payload) as r:
        # dg_performance_total_ms = r.headers.get('x-dg-performance-total-ms', 'Not Available')
        # print(f"Deepgram Performance Total (ms): {dg_performance_total_ms}ms")
        print(f"request: {r}")

        for chunk in r.iter_content(chunk_size=1024):
              # do something

However I got a 400 error for the request:

{"err_code":"INVALID_QUERY_PARAMETER","err_msg":"Failed to deserialize query parameters: provided string was not `true` or `false`","request_id":"0b23a77e-4096-4861-83e1-d8855075fecd"}'

Thanks for any help.

Regards,
Jonathan

My environment:

• Python 3.11.10 (main, Sep 28 2024, 19:10:46) [Clang 11.0.3 (clang-1103.0.32.62)] on darwin
• deepgram-sdk 3.7.2
• mac os Catalina 10.15.7
• Deepgram plan: Pay As You Go

[jonathan885]

found my issues

• I was using the https://api.beta.deepgram.com/v1/speak endpoint.
• the payload has an extra key: voice not recognized by the endpoint
• performance_rate is not a query parameter for the regular one (https://api.deepgram.com/v1/speak?model)

So the working request (in my case) is:

MODEL_NAME = "aura-asteria-en"
DEEPGRAM_URL = f"https://api.deepgram.com/v1/speak?model={MODEL_NAME}&encoding=linear16&sample_rate=24000"
headers = {
        "Authorization": f"Token {config['DEEPGRAM_KEY']}",
        "Content-Type": "application/json"
    }

    payload = {
        "text": text
    }
with requests.post(DEEPGRAM_URL, stream=True, headers=headers, json=payload) as r:
    ...