Arcx Bing Speech Recognition Support

ARCx does not use the old Bing Speech Recognition robot skill as a standalone skill in the same way ARC did.

What to use in ARCx instead

The closest equivalents in ARCx are:

• Azure Speech To Text Engine
• Speech Phrase Recognition
• Windows Speech Engine (Windows only)
• SherpaOnnx Speech To Text

For a setup that behaves most like the old Bing command workflow, I’d recommend:

1. Add Azure Speech To Text Engine
2. Add Speech Phrase Recognition
3. Put your command phrases in the phrase table
4. Use the phrase scripts to run your actions

That’s the standard ARCx approach:

• the engine turns speech into text
• Speech Phrase Recognition matches the text to commands

Why this is the recommended path

In ARCx, speech is split into separate skills instead of one all-in-one skill. That gives you more flexibility, but it also means the older Bing-style skill isn’t the preferred architecture anymore.

If Bing was your most reliable option, Azure Speech To Text Engine is likely the closest replacement to try first:

• it uses Synthiam’s service
• no Azure key is required
• it works well for one-phrase-at-a-time recognition

Docs:

• Azure Speech To Text Engine: https://synthiam.com/Support/Skills/Audio/Azure-Speech-To-Text-Engine?id=23502
• Speech Phrase Recognition: https://synthiam.com/Support/Skills/Audio/Speech-Phrase-Recognition?id=22302
• How to set up speech recognition in ARCx: https://synthiam.com/Support/using-arcx/speech-recognition/how-to-setup-speech-recognition

About plans for Bing in ARCx

I can’t confirm any public roadmap commitment to bring back the legacy Bing skill as-is. The current ARCx direction is to use the newer speech stack above rather than the older ARC skill model.

Recommended setup if you want Bing-like reliability

If your goal is reliable phrase triggering, I’d try this order:

1. Azure Speech To Text Engine
2. Speech Phrase Recognition
3. Keep phrases short and distinct, and prefix them with the robot name if needed

Example:

• robot stop
• robot forward
• robot turn left

If you want, I can help you build the exact ARCx replacement

If you tell me:

• which microphone you’re using,
• whether you want always listening or push-to-talk,
• and a few sample phrases,

I can recommend the best ARCx speech setup for your robot.

Questions like this are often best handled in ARC’s built-in AI Assistant too, since it can help rewrite or organize your speech scripts right in the editor: https://synthiam.com/Products/AI-Assistant

Since you're using a wake word. Don't use that old cloud-based stuff. Use the localized Onyx speech recognition with a configured wake word.

Don't try to convert an entire project over to ARCx. You will have too much to learn, and it will be frustrating. Instead, tackle one thing at a time. Just start with a new basic empty project and play with various ways to use speech recognition the way you like it.

Also, i should add that you have been using Azure Speech Services since January 14, 2020. The robot skill has continued to be named "bing" as to not confuse people who were used to it The Bing Speech API was officially retired on January 14, 2020. Microsoft deprecated this service to consolidate its cloud offerings into Azure Cognitive Services. [1]

• The Replacement: Developers were required to migrate their applications to the unified Azure Speech Services and the Speech SDK. This replacement provides speech-to-text (speech recognition), translation, and voice intent capabilities. [1]

Exactly - that’s the key distinction.

The old Bing Speech Recognition name was retained in ARC for familiarity, but under the hood it has been using Azure Speech Services for years. So if you were relying on that skill, what you really valued was the Azure-backed speech recognition behavior, not the retired Bing API itself.

In ARCx, the better path is to start fresh

DJ’s advice is the right approach:

• Don’t try to "lift and shift" the whole ARC project
• Start with a new empty ARCx project
• Add one speech workflow at a time
• Test the behavior you want before rebuilding the rest of the project

That will save you a lot of frustration, especially because ARCx’s speech architecture is different.

Best ARCx speech options to try

If you want localized speech recognition with a wake word, I’d suggest this order:

Option 1: SherpaOnnx Speech To Text

• Local/offline
• Good for always-on wake word or command detection
• Best if you want to avoid cloud services

Option 2: Windows Speech Engine

• Windows only
• Also local
• Very reliable for command phrases on supported systems

Option 3: Azure Speech To Text Engine

• Cloud-based like the old "Bing" behavior
• No Azure key required in the Synthiam version
• Good if you still want cloud accuracy

Then pair it with:

• Speech Phrase Recognition for command matching
  https://synthiam.com/Support/Skills/Audio/Speech-Phrase-Recognition?id=22302

Why this matters

ARCx separates speech into pieces:

• Microphone
• Speech-to-text engine
• Phrase recognition
• Your scripts

That’s different from the older all-in-one pattern, so rebuilding one piece at a time is the proven way to get it working cleanly.

Suggested first test

If your goal is just "wake word then commands," try this simple ARCx setup:

1. Create a new empty project
2. Add SherpaOnnx Speech To Text
3. Add Speech Phrase Recognition
4. Put in a few distinct phrases like:
   • robot stop
   • robot forward
   • robot left
5. Test whether the engine hears the wake word or command phrases reliably
6. Then expand from there

ARCx speech setup guide: https://synthiam.com/Support/using-arcx/speech-recognition/how-to-setup-speech-recognition

Bottom line

You’re not really losing the old capability - you’re just moving to the modern ARCx architecture, where the equivalent behavior is built from newer components.

If you want, I can help you map your current ARC Bing Speech Recognition setup into the closest ARCx speech workflow step by step.

What are you ultimately trying to do with speech on the robot - wake word, command phrases, or free-form conversation?

Fascinating. Thanks for the sage advice! I really appreciate the guidance in explanations.

Resolved by the experts. Happy times indeed!