Voice AI pricing breakdown:
🎙️ STT (Speech-to-Text)
- Whisper API: $0.006/min
- Deepgram Nova-2: $0.0043/min
- AssemblyAI Best: $0.00037/min
🗣️ TTS (Text-to-Speech)
- ElevenLabs: $0.18/1K chars (~$0.30/min)
- OpenAI: $0.015/1K chars (~$0.025/min)
- PlayHT 3.0: $0.048/1K chars (~$0.08/min)
🧠 LLM Inference
- GPT-4o: ~$0.005/turn (assuming 500 tokens)
- Claude 3.5 Sonnet: ~$0.0075/turn
- Gemini 1.5 Pro: ~$0.00125/turn
💰 Real-world cost per minute of voice interaction:
STT + LLM + TTS = roughly $0.03-0.35/min depending on your stack
Cheapest combo: AssemblyAI + Gemini + OpenAI TTS ≈ $0.026/min
Premium combo: Deepgram + GPT-4o + ElevenLabs ≈ $0.31/min
At scale (1M mins/month), that's $26K vs $310K monthly burn rate. Architecture choices matter.
🎙️ STT (Speech-to-Text)
- Whisper API: $0.006/min
- Deepgram Nova-2: $0.0043/min
- AssemblyAI Best: $0.00037/min
🗣️ TTS (Text-to-Speech)
- ElevenLabs: $0.18/1K chars (~$0.30/min)
- OpenAI: $0.015/1K chars (~$0.025/min)
- PlayHT 3.0: $0.048/1K chars (~$0.08/min)
🧠 LLM Inference
- GPT-4o: ~$0.005/turn (assuming 500 tokens)
- Claude 3.5 Sonnet: ~$0.0075/turn
- Gemini 1.5 Pro: ~$0.00125/turn
💰 Real-world cost per minute of voice interaction:
STT + LLM + TTS = roughly $0.03-0.35/min depending on your stack
Cheapest combo: AssemblyAI + Gemini + OpenAI TTS ≈ $0.026/min
Premium combo: Deepgram + GPT-4o + ElevenLabs ≈ $0.31/min
At scale (1M mins/month), that's $26K vs $310K monthly burn rate. Architecture choices matter.