The Problem with TTS Today
Most text-to-speech APIs fall into two camps:
- Expensive but good: ElevenLabs, Azure Neural TTS - great quality but $0.15-0.30/min
- Cheap but limited: edge-tts, Google Cloud TTS - free/cheap but no voice customization
What if you could generate any voice you wanted from a text description alone - at 1/6th the cost of ElevenLabs?
That's what we built with VoxCPM2 on PixelAPI.
What We Built
A self-hosted VoxCPM2 TTS API with two modes:
Voice Design: Generate speech using only a text description. No reference audio needed.
(A warm elderly man, gentle voice) Hello, welcome to our store.
Voice Cloning: Upload a reference audio clip and clone the voice for consistent brand voices.
Voice Design Quality - Real Benchmarks
We tested 8 different voice profiles using vision QC on spectrograms:
| Voice | Score | Notes |
|---|---|---|
| Young Woman (gentle) | 8/10 | Clear formants, good dynamics |
| Child (curious) | 8/10 | Good high-end presence, no artifacts |
| Villain (dark) | 7/10 | Minor HF noise, good menace |
| Elderly Man (wise) | 7/10 | Slightly weak volume but clean |
Speed on Real Hardware
Benchmarked on RTX 6000 Ada:
| Audio Duration | Generation Time | RTF |
|---|---|---|
| 10 seconds | ~4 seconds | 0.38 |
| 25 seconds | ~9 seconds | 0.38 |
RTF of 0.38 means it generates faster than real-time.
Pricing - The 2x Cheaper Rule
| Provider | Voice Design | Voice Cloning |
|---|---|---|
| ElevenLabs | $0.30/min | $0.30/min + subscription |
| Play.ht | $0.036/min | $0.10/min |
| PixelAPI | $0.050/min | $0.100/min |
6x cheaper than ElevenLabs for voice design, and we're self-hosted.
API Example - Python
import requests
resp = requests.post("https://api.pixelapi.dev/v1/tts/generate",
headers={"Authorization": "Bearer YOUR_KEY"},
data={"text": "(A warm elderly man, gentle) Hello everyone.", "language": "en"}
)
job = resp.json()
print(f"Job: {job['id']}")
30 Languages Supported
VoxCPM2 supports 30 languages including English, Hindi, Chinese, Spanish, French, German, Japanese, Korean, Russian, Arabic, and more.
The API is live at api.pixelapi.dev with 100 free credits on signup.
Top comments (0)