Synthesize Chinese, the sound of the last kanji lasts for a long time #3049
Replies: 3 comments 2 replies
-
|
tts.tts_to_file(text="北京欢迎你。不到长城非好汉 北京是六朝古都。", file_path="c3.wav")，need 。at the end |
Beta Was this translation helpful? Give feedback.
-
|
注意下标点试试 |
Beta Was this translation helpful? Give feedback.
-
|
你好，可以请教几个关于tts 中文模型的问题吗？有哪些已经训练好的中文模型的效果较好？如何调优？望回复 |
Beta Was this translation helpful? Give feedback.
-
|
hi 求教下，你阅读了相关代码了吗，我不太明白，为什么推理过程分为一下两步 gpt_codes = gpt_inference.generate(text, cond_latents) 为什么先GPT生成离散的 codes，再跑一次GPT生成离线的float的 latents 相关问题我也放在这里了 #3223 |
Beta Was this translation helpful? Give feedback.
-
|
This design comes from Tortoise, the author indicates latents got richer semantic info than codes |
Beta Was this translation helpful? Give feedback.
-
tts.tts_to_file(text="北京欢迎你。不到长城非好汉 北京是六朝古都", file_path="c3.wav")
the pronounce of ‘北京是六朝古都’ repeat in c3.wav
Beta Was this translation helpful? Give feedback.
All reactions