speech model
OpenAI TTS
OpenAISteerable voice — you describe the delivery in a prompt rather than tuning sliders.
OpenAI’s TTS models are less expressive than ElevenLabs at the top end but introduce a genuinely useful idea: you instruct the delivery in natural language ("warm, unhurried, like you are explaining to a nervous candidate"). For screening flows where tone should shift with context, that is easier to control than a fixed voice preset.
CapabilitiesPrompted deliveryStreamingLow costPreset voices
Variants
2 ways to run it.
Same family, different quality/cost rung. Picking the wrong one is the most common way teams overpay.
Mini TTS
Balancedgpt-4o-mini-ttsSteerable delivery at low cost.
400ms
latency
1×
rel. cost
82
quality
TTS-1 HD
Hightts-1-hdHigher-fidelity preset voices, no steering.
700ms
latency
2×
rel. cost
78
quality
Fit
Two audiences, one model.
The same model is a different proposition depending on what you point it at.
76/100
Recruitment fit
The cost-effective choice when tone needs to vary by stage but the voice does not need to be anyone real.
Best for
- Automated candidate updates
- Interview instructions
- High-volume notifications
Strengths
- Prompted delivery adapts tone per stage without swapping voices
- Materially cheaper than ElevenLabs at volume
- No cloning means no consent problem
Watch out for
- Less expressive at the top end
- No custom voice cloning
