OpenAI TTS

from deepeval.models import OpenAITTSModel
from deepeval.voice import VoiceConfig

tts_model = OpenAITTSModel(
    model="gpt-4o-mini-tts",
    voice="coral",
)
voice_config = VoiceConfig(
    tts_model=tts_model,
    ...,
)

There are ZERO mandatory and SEVEN optional parameters when creating an OpenAITTSModel:

• [Optional] model: A string specifying the name of the speech model to use. Defaulted to gpt-4o-mini-tts.
• [Optional] api_key: A string specifying your OpenAI API key. Defaults to OPENAI_API_KEY if not passed; raises an error at runtime if neither is set.
• [Optional] base_url: A string specifying a custom endpoint to reach the OpenAI API through. Defaulted to None.
• [Optional] voice: A string naming the OpenAI voice to speak in, such as "alloy", "onyx", or "shimmer". Defaulted to alloy.
• [Optional] response_format: A string audio format for synthesized turns — one of "wav", "mp3", "opus", "aac", "flac", or "pcm". Defaulted to "wav".
• [Optional] cost_per_1m_chars: A float overriding the per-character price used for cost accounting. Defaults to the built-in price for the model.
• [Optional] generation_kwargs: A dictionary of additional parameters forwarded to OpenAI's audio.speech.create(...) call.

First set your API key:

# .env.local
OPENAI_API_KEY=<your-openai-api-key>
DEEPEVAL_TTS_MODEL=tts-1-hd

OpenAI is the speech provider when no USE_*_TTS flag is set, so USE_OPENAI_TTS=1 is only needed to switch back after selecting another provider. deepeval set-tts openai --model tts-1-hd writes both for you.

You can also pass the model name in code, which wins over DEEPEVAL_TTS_MODEL:

from deepeval.voice import VoiceConfig

voice_config = VoiceConfig(
    tts_model="tts-1-hd",
    ...,
)

A bare model name is shorthand for OpenAITTSModel(model=...), so you can leave tts_model out entirely and synthesis still runs on gpt-4o-mini-tts. Pass the model object itself to pick a voice.