For a real-time usage scenario, updating the text if incorrect speech is detected might not be feasible. But if you are planning to use the speech translation service to translate a batch of audio files and then include a review process then it is possible to correct the text and then pass them to TTS service to generate a voice with translated text.
You can use the subscription keys in more than one app at the same time but the usage limits still apply on the resource. So, if one of your app has used up the TPS limits then the subsequent call might fail. You can however use standard pricing tier to get higher TPS for such scenarios. It is recommended though to use separate resources for separate applications to track the requests. Thanks!!
If an answer is helpful, please click on or upvote which might help other community members reading this thread.