moebiussurfing
commented
1 month ago I tried to bypass the openAI account as explained here https://github.com/lamm-mit/PDF2Audio/issues/1
but only works for the first phase, submitting the pdf to generate the podcast text. When trying to generate the audio voice the openAI key error remains:
RAW RESPONSE:
{"id": "chatcmpl-37039926-333d-4250-ac91-e1c4a1046396", "choices": [{"finish_reason": "stop", "index": 0, "logprobs": null, "message": {"content": " \n\n<assistant>\n```json\n{\n \"scratchpad\": \"During our research, we've found that AI is being used to identify and manipulate our emotions. We need to understand how it works so we can create boundaries and protect ourselves from
...
Logging Details LiteLLM-Success Call: Cache_hit=None
Looking up model=custom_model in model_cost_map
Looking up model=custom_model in model_cost_map
2024-09-30 20:46:34.907 | INFO | __main__:generate_audio:623 - Dialogue generation completed.
2024-09-30 20:46:34.908 | INFO | __main__:generate_audio:630 - Generating audio...
2024-09-30 20:46:35.186 | ERROR | __main__:generate_audio:668 - An error occurred in generate_audio: Error code: 401 - {'error': {'message': 'Incorrect API key provided: DUMMY. You can find your API key at https://platform.openai.com/account/api-keys.', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_api_key'}}
Traceback (most recent call last):
How about adding Text-to-Speech alternatives to openai, such as: deepgram, fish.audio. Similarly adding other LLMs as well.