One Azure key for listening and speaking
Dograh is the self-hostable voice agent platform with an open-source workflow builder. Connect your Azure Speech key and region once, and the same resource transcribes the caller and voices the agent, self-hosted or on Dograh Cloud.
Connect once, use it twice
Add Azure Speech under Providers
Regionwhere your Azure resource runs
Click Add Provider, choose Azure Speech Services and enter your Speech key and Region. East US is used when no region is set.
→ Create the resource in the region your data must stay in
One connection, two tabs
Localethe language the recognizer expects
In a BYOK model configuration, pick the Azure connection on the STT tab and again on the TTS tab. en-US is the default locale.
→ Pick the locale your callers actually speak
Choose a voice, then call it
Call tracethe timing of every turn
Voice defaults to en-US-AriaNeural and takes a typed voice name. The Phone Call button rings a number you enter, and the trace shows each turn's timing.
→ Judge two or three voices on real calls
Azure listens and speaks, the rest is yours
Microsoft stores no real-time audio at rest, and each turn still travels to your Azure region
Audio streams at the line's own rate, 8 kHz on Twilio and 16 kHz on Vonage
Recordings and transcripts sit in the platform's own storage, on your servers when you self-host
Azure where you want it, a platform you own
One subscription, three layers
Speech to text, the voice and an Azure OpenAI language model can all bill to the Azure account you already hold.
The region you choose
The resource's region decides where transcription and voice run, so a West Europe resource keeps both in West Europe.
Open platform code
Dograh is open source under BSD-2-Clause, so your team can read and change every part of the agent platform.
No per-minute platform fee self-hosted
Servers, model keys and phone minutes still cost money. Dograh Cloud charges a 1¢-a-minute platform fee.
Check these before the agent takes calls
- Region chosenThe Speech resource sits in the region your data has to stay in.
- Locale matchedThe speech-to-text locale matches the language your callers speak.
- Voice heard on a callTwo or three voices were judged over a real phone line, not only in a browser.
- Transfer tested by phoneA transfer was tried on a real phone call, since browser test calls cannot transfer.
Can one Azure key cover speech-to-text and the voice?+×
Yes. The same Speech key and region serve both. Add the connection once, then pick it on the STT and TTS tabs.
Does Azure Speech handle 8 kHz phone audio?+×
Yes. Azure's recognizer accepts 8 kHz audio, and Dograh asks Azure for 8 kHz voice audio, the rate most phone lines carry.
Which Azure voice does Dograh use by default?+×
en-US-AriaNeural, a standard Neural voice. The Voice field also accepts a typed voice name, with Speed from 0.5 to 2.0.
Can I use Azure OpenAI as the language model?+×
Yes. The LLM tab can point at Azure OpenAI, so speech to text, the voice and the language model sit in one subscription.
Is Azure Voice Live the same thing?+×
No. Voice Live is Microsoft's separate speech-to-speech API. Dograh uses Azure Speech for transcription and voice, and Azure OpenAI Realtime in Realtime mode.
Who pays for the Azure usage?+×
Azure bills your own account. Self-hosted Dograh has no per-minute platform fee, though servers, model keys and phone minutes cost money. Dograh Cloud charges 1¢ a minute.
Add your Speech key and ring the agent.
Self-host in one command, or start on Dograh Cloud.
$ curl … start_docker.sh && ./start_docker.sh