@ -138,7 +138,7 @@ You can check the text of downloaded wavs in `source/README.md`.
```bash
```bash
./run.sh --stage 3 --stop-stage 3 --gpus 0
./run.sh --stage 3 --stop-stage 3 --gpus 0
```
```
`stage 3` of `run.sh` calls `local/synthesize_e2e.sh`, `stage 0` of it is **Speech Synthesis** and `stage 1` of it is **Speech Editing**.
`stage 3` of `run.sh` calls `local/synthesize_e2e.sh`. `synthesize_e2e.sh` is a script for performing both **Speech Synthesis** and **Speech Editing** tasks by default. It converts input text into speech for synthesis and modifies existing speech based on new text content for editing.
You can modify `--wav_path`、`--old_str` and `--new_str` yourself, `--old_str` should be the text corresponding to the audio of `--wav_path`, `--new_str` should be designed according to `--task_name`, both `--source_lang` and `--target_lang` should be `zh` for model trained with AISHELL3 dataset.
You can modify `--wav_path`、`--old_str` and `--new_str` yourself, `--old_str` should be the text corresponding to the audio of `--wav_path`, `--new_str` should be designed according to `--task_name`, both `--source_lang` and `--target_lang` should be `zh` for model trained with AISHELL3 dataset.