Instructions to use hetanshwaghela/amnesiac-voxcpm-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- VoxCPM
How to use hetanshwaghela/amnesiac-voxcpm-lora with VoxCPM:
import soundfile as sf from voxcpm import VoxCPM model = VoxCPM.from_pretrained("hetanshwaghela/amnesiac-voxcpm-lora") wav = model.generate( text="VoxCPM is an innovative end-to-end TTS model from ModelBest, designed to generate highly expressive speech.", prompt_wav_path=None, # optional: path to a prompt speech for voice cloning prompt_text=None, # optional: reference text cfg_value=2.0, # LM guidance on LocDiT, higher for better adherence to the prompt, but maybe worse inference_timesteps=10, # LocDiT inference timesteps, higher for better result, lower for fast speed normalize=True, # enable external TN tool denoise=True, # enable external Denoise tool retry_badcase=True, # enable retrying mode for some bad cases (unstoppable) retry_badcase_max_times=3, # maximum retrying times retry_badcase_ratio_threshold=6.0, # maximum length restriction for bad case detection (simple but effective), it could be adjusted for slow pace speech ) sf.write("output.wav", wav, 16000) print("saved: output.wav") - Notebooks
- Google Colab
- Kaggle
Publish verified A.M.N. adapter from step_0000090
Browse files- README.md +20 -0
- lora_config.json +28 -0
- lora_weights.safetensors +3 -0
README.md
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: openbmb/VoxCPM1.5
|
| 3 |
+
library_name: voxcpm
|
| 4 |
+
pipeline_tag: text-to-speech
|
| 5 |
+
tags:
|
| 6 |
+
- lora
|
| 7 |
+
- voxcpm
|
| 8 |
+
- text-to-speech
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# A.M.N. VoxCPM1.5 LoRA
|
| 12 |
+
|
| 13 |
+
Single-speaker LoRA adapter selected from checkpoint step 90 after human listening on 110 curated A.M.N. voice clips.
|
| 14 |
+
|
| 15 |
+
- Base model: `openbmb/VoxCPM1.5`
|
| 16 |
+
- LoRA: LM + DiT, rank 16, alpha 32
|
| 17 |
+
- Native output: 44.1 kHz
|
| 18 |
+
- Intended use: the AMNESIAC fictional interrogator voice
|
| 19 |
+
|
| 20 |
+
Synthetic speech should be disclosed where appropriate. Do not use this adapter for impersonation, fraud, or disinformation.
|
lora_config.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_model": "openbmb/VoxCPM1.5",
|
| 3 |
+
"lora_config": {
|
| 4 |
+
"enable_lm": true,
|
| 5 |
+
"enable_dit": true,
|
| 6 |
+
"enable_proj": false,
|
| 7 |
+
"r": 16,
|
| 8 |
+
"alpha": 32,
|
| 9 |
+
"dropout": 0.0,
|
| 10 |
+
"target_modules_lm": [
|
| 11 |
+
"q_proj",
|
| 12 |
+
"v_proj",
|
| 13 |
+
"k_proj",
|
| 14 |
+
"o_proj"
|
| 15 |
+
],
|
| 16 |
+
"target_modules_dit": [
|
| 17 |
+
"q_proj",
|
| 18 |
+
"v_proj",
|
| 19 |
+
"k_proj",
|
| 20 |
+
"o_proj"
|
| 21 |
+
],
|
| 22 |
+
"target_proj_modules": [
|
| 23 |
+
"enc_to_lm_proj",
|
| 24 |
+
"lm_to_dit_proj",
|
| 25 |
+
"res_to_dit_proj"
|
| 26 |
+
]
|
| 27 |
+
}
|
| 28 |
+
}
|
lora_weights.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dddfa755e9439f1477cb7f1a29daa2dd29b1f577391a18567736532403386e83
|
| 3 |
+
size 16421176
|