The objective of this project is to generate a podcast episode based on the content of a given article. This is achieved using entirely open-source large language models (LLMs), avoiding reliance on expensive proprietary AI platforms.
Our tech stack includes:
- Ollama: A local LLM runtime for generating conversational scripts.
- ElevenLabs: A text-to-speech platform for converting scripts into audio.
To run Ollama effectively, your system should meet the following minimum requirements:
-
Operating System: Ubuntu 22.04 or later
-
CPU: Modern Intel or AMD processor with at least 4 cores. For models up to 13B parameters, an 8-core CPU is recommended
-
RAM:
- 7B models: At least 8GB
- 13B models: At least 16GB
-
Disk Space: Approximately 12GB for installing Ollama and basic models. Additional space is required for storing model data depending on the models used
-
GPU: Optional. While not required, a GPU can improve performance, especially when working with large models
ElevenLabs offers various subscription plans suitable for different needs:
-
Free Plan:
- 10,000 credits per month
- Approximately 10 minutes of high-quality text-to-speech
- $0 per month
-
Starter Plan:
- 30,000 credits per month
- Approximately 30 minutes of high-quality text-to-speech
- $5 per month
- Includes additional features like commercial license and instant voice cloning
For basic podcast generation, the Free Plan may suffice. However, for extended usage and additional features, the Starter Plan is recommended.
- Select an Article: Choose a web article that permits content scraping.
- Generate a Script: Use the LLM to create a conversation between two fictional podcast hosts based on the article's content.
- Produce the Audio: Use ElevenLabs to synthesize the script into a realistic podcast dialogue.