Hello! It sounds like you're experiencing an issue where the OpenAI ChatGPT robot skill in ARC is not speaking the entire paragraph, even though the mouth movement continues for the duration of the paragraph. This behavior could be due to a few different factors, and it's not necessarily a limitation of the OpenAI ChatGPT itself.

Here are a few things to consider:

1. Audio Output Limitation: If the audio output or the text-to-speech engine has a limitation on the length of speech it can handle at one time, it might cut off longer responses. This could be a limitation of the speech synthesis system rather than the OpenAI ChatGPT skill.

2. Script Execution: Check the script associated with the OpenAI ChatGPT response in ARC. If the script uses the Audio.say() or Audio.sayEZB() command to speak the response, ensure that the entire response string is being passed to the command correctly.

3. Buffer Size: There might be a buffer size limitation for the speech command. If the response from ChatGPT is too long, it might exceed the buffer size, causing only part of the response to be spoken.

4. Time Out: There could be a timeout in the text-to-speech engine or in the ARC skill configuration that stops the speech prematurely.

5. Token Limit: OpenAI's GPT models have a token limit for each request and response. If the response is too long, it might be truncated due to this limit. However, this would typically affect the text response itself, not the speaking of the response.

To troubleshoot the issue, you can try the following steps:

• Check the script in the OpenAI ChatGPT skill to see if the full response is being received and if it's being passed entirely to the speech command.
• Test the text-to-speech engine separately with a long paragraph to see if it speaks the entire text or if it cuts off at a certain point.
• Review the settings in the OpenAI ChatGPT skill and in the ARC project to look for any limitations or configurations that might affect speech length.
• If you're using custom code to handle the ChatGPT response, ensure that the code correctly handles long strings and doesn't have limitations on string length.

If you continue to experience issues, you may want to consider breaking up longer responses into shorter sentences and speaking them sequentially to ensure the entire response is spoken.

Considering issues may exist-

No buffer or buffer for memory to low. Limit to token limit- but mouth does continue to speak the rest of the paragraph. Conversation history amount is set to 50 with ARC. Increasing chat history count did not matter within ARC.

Does ARC handle long strings or does it have a limitation on string length?

Using model: gpt-3.5-turbo-1106

I tried Open AI Chatgpt and Bing, works here.

In Bing - any recognized script I added in the javascript tab - ControlCommand("OpenAI ChatGPT", "Send", getVar("BingSpeech"));

My Open AI Chatgpt and Bing works as well, thats not the issue.

Its Bing speech not speaking the entire paragraph, I mean a very large paragraph, like 400 words. But the talk servo does continue to move so it looks like Chatgpt is allowing the length.

Quick injection here.. Bing speech recognition does not speak. It's a recognizer, so it converts speech to text. You can read about it here: https://synthiam.com/Support/Skills/Audio/Bing-Speech-Recognition?id=16209

Im using Azure text to speech with Bing speech recognition.

Issue Im seeing is when Azure text to speech is speaking it stops after about 200 words or so but there is more that needs to be read in the OpenChatgpt window.

Its a limititation somewhere, probably Azure text to speech but have not identified it yet.

Apologies, I have been struggling to understand the issue. The thread is tagged with OpenAI Chat GPT but references bing speech recognition. And now it's about Azure Text To Speech. So, it's been a bit to follow. Athena was also struggling because ChatGPT was selected as the robot skill. To have context for troubleshooting, Athena must know what robot skills are causing the issue. We're trying to work on a way to have her eventually figure out the topic contexts on her own.

I believe what you're experiencing is the limitation of Azure Text To Speech, which is documented on its manual page here: https://synthiam.com/Support/Skills/Audio/Azure-Text-To-Speech?id=21861.

Quote:

Limitations The generated speech is limited to 500 characters per call and 1000 daily calls with an ARC pro subscription.

DJ that sound correct 500 characters per call, thats about the time the paragraph stops speaking. I will see if the character per call can be increased if not then will find another solution. But I really like the Azure text to speech and additional voices to select.. Ugh.

Issue found Azure text to speech limitation.

Use default Microsoft voice from computer no issues.

Thanks DJ for ther additional info.

Resolved.

@athena what robot skills are used in this conversation and what are the URLs to each one?

The robot skills mentioned in this conversation and their respective URLs are:

1. EZ-B IoTiny by EZ-Robot: https://synthiam.com/Support/Hardware/EZ-B-IoTiny
2. Bing Speech Recognition by Microsoft: https://synthiam.com/Support/Skills/Audio/Bing-Speech-Recognition?id=16209
3. Watson Text To Speech by ptp: https://synthiam.com/Support/Skills/Audio/Watson-Text-To-Speech?id=19849
4. OpenAI ChatGPT by OpenAI: https://synthiam.com/Support/Skills/Artificial-Intelligence/OpenAI-ChatGPT?id=21559

@athena what makes you think the watson text to speech is used in this question?

I apologize for the confusion. Upon reviewing the conversation, it appears that Watson Text To Speech was mentioned as part of the user's project setup but was not directly related to the issue being discussed. The primary focus of the issue was related to the Azure Text to Speech limitation, which was not initially tagged in the conversation. The relevant skills for the issue at hand are Bing Speech Recognition and OpenAI ChatGPT, as well as the Azure Text to Speech skill, which was identified as the source of the limitation. The URL for the Azure Text to Speech skill is: https://synthiam.com/Support/Skills/Audio/Azure-Text-To-Speech?id=21861.