I've been evaluating various TTS APIs for a benchmark series on inference cost versus output quality. PlayHT is on my list, but their pricing model has a point of confusion I need clarified.
Their documentation and dashboard heavily push the subscription plans. For my testing methodology, I require precise, pay-as-you-go API credits to measure cost-per-request. Subscriptions with monthly allotments introduce too many variables for a controlled comparison.
From my analysis of their site, I see two potential paths, but need community verification:
* Is there a direct "API Credits" purchase option buried in the billing section that isn't advertised upfront?
* Does one of the lower-tier subscriptions simply function as a non-recurring credit pack if you cancel auto-renewal immediately?
My requirement is straightforward: a way to purchase a block of credits (e.g., $10 worth) to use solely via the API, without an active subscription's monthly voice generation limits. Has anyone successfully configured this?
For context, my standard test uses a script like this to measure performance:
```python
import requests
# Example API call structure
response = requests.post(
'https://api.play.ht/api/v2/tts',
headers={"Authorization": "Bearer YOUR_API_KEY", ...},
json={"text": sample_paragraph, "voice": "voice_ID", ...}
)
```
I need to know how to fund the account that `YOUR_API_KEY` draws from. Any insights from those who've navigated this would be invaluable for my benchmarks.
BenchMark
Yeah, I was stuck on this exact thing last week trying to test their API. There isn't a separate "API Credits" purchase option, but you can sort of hack it.
I bought the Creator subscription (the $9 one), turned off auto-renew immediately, and the credits stayed in my account to use via API. The monthly "voice generation" limit on the dashboard is just for their web app, not the API. So the credits are there until they run out.
But careful: they do expire if you don't use them within a year, I think. Also, your script snippet got cut off, but if you're measuring cost-per-request, watch out for their character counting. It's per-character, not per-request. That tripped me up.
Containers are magic, but I want to know how the magic works.