Commit
•
d9f52f7
1
Parent(s):
b8d92b6
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,4 +1,6 @@
|
|
| 1 |
This is a finetune of the base tortoise model, it was trained on 680 hours of dutch (flemish accent) speech, mostly from audiobooks.
|
| 2 |
The one downside is that you can't use the zero shot capability of the model, it needs to be finetuned for a specific voice. But once you do that the results are great.
|
| 3 |
|
| 4 |
-
You can use it with the [mrq webui](https://git.ecker.tech/mrq/ai-voice-cloning), be sure to read the wiki.
|
|
|
|
|
|
|
|
|
| 1 |
This is a finetune of the base tortoise model, it was trained on 680 hours of dutch (flemish accent) speech, mostly from audiobooks.
|
| 2 |
The one downside is that you can't use the zero shot capability of the model, it needs to be finetuned for a specific voice. But once you do that the results are great.
|
| 3 |
|
| 4 |
+
You can use it with the [mrq webui](https://git.ecker.tech/mrq/ai-voice-cloning), be sure to read the wiki. The model only works if you use the dutch_vl.json tokenizer.
|
| 5 |
+
|
| 6 |
+
There are multiple checkpoints, the number in the name is the number of steps the model trained on, higher is probably better.
|