CSM is a speech generation model from Sesame,
[SesameAILabs/csm](https://github.com/SesameAILabs/csm).
It can generate a conversational speech between two different speakers.
The speakers turn are delimited by the `|` character in the prompt.
```bash
cargo run --example csm --features cuda -r -- \
```