Dialogflow CX voice agents use Speech-to-Text for speech recognition, which is included in Dialogflow CX pricing. Dialogflow CX automatically selects a speech recognition model for you, but you can optionally specify the model.
Available models
All available models are listed at Speech-to-Text models. Select a model that is best suited to your domain and supports your agent language and speech features.
If a model is not explicitly specified, then Dialogflow CX auto-selects a model based on the audio configuration in API requests and agent settings.
The following models typically have the best performance and broad regional serving capacity:
telephony_short(best for telephony Dialogflow CX)telephony(best for Agent Assist) (also good for telephony Dialogflow CX when advanced timeout-based end of speech sensitivity is enabled)phone_call(good for Agent Assist and telephony Dialogflow CX)latest_short(best for non-telephony Dialogflow CX)command_and_search(best for languages where other models aren't available)
To select Speech-to-Text V2 (chirp_3) in the Dialogflow CX console, your
project requires access approval. High-volume production deployments also
require an advance capacity review with your Google Cloud account team.
Dialogflow CX agents in us-central1 route Speech-to-Text V2 traffic to the us
multi-region endpoint, except for chirp_2 requests. The single-region chirp
model isn't available on that endpoint. For 8 kHz MULAW telephony audio, use
chirp_3, telephony, or telephony_short instead of the broadband
Speech-to-Text V2 short model.
Specify a model
You can supply the model for an agent, flow, or page with the model selection setting.You can also supply the model when calling the
Sessions.detectIntent or Sessions.streamingDetectIntent methods;
Select a protocol and version for the Session reference:
| Protocol | V3 | V3beta1 |
|---|---|---|
| REST | Session resource | Session resource |
| RPC | Session interface | Session interface |
| C++ | SessionsClient | Not available |
| C# | SessionsClient | Not available |
| Go | SessionsClient | Not available |
| Java | SessionsClient | SessionsClient |
| Node.js | SessionsClient | SessionsClient |
| PHP | Not available | Not available |
| Python | SessionsClient | SessionsClient |
| Ruby | Not available | Not available |
ConversationProfile
for
Agent Assist.
Specifying the model in a detect intent or conversation profile API call will
override any model selections applied to the agent, flow, or page, unless you
enable the
Override request-level speech model
setting.