FR_M1 mixed excitation: less buzz (same weights, new vocoder setting, ONNX and samples)
Browse files- README.md +5 -2
- config.json +12 -2
- onnx/model.onnx +2 -2
- samples/FR_M1-1.wav +1 -1
- samples/FR_M1-2.wav +1 -1
- samples/FR_M1-3.wav +1 -1
- samples/FR_M1-4.wav +1 -1
- tiny-voca-fr.pt +2 -2
README.md
CHANGED
|
@@ -67,8 +67,11 @@ sentences.
|
|
| 67 |
|
| 68 |
## Limits
|
| 69 |
|
| 70 |
-
- The male voice
|
| 71 |
-
|
|
|
|
|
|
|
|
|
|
| 72 |
- eSpeak NG misses some liaisons (*comment‿allez*), adds a few false ones, and
|
| 73 |
reads *fils* (son) as [fil].
|
| 74 |
- Two fixed voices. No voice cloning.
|
|
|
|
| 67 |
|
| 68 |
## Limits
|
| 69 |
|
| 70 |
+
- The male voice is less natural than the female voice. Its vocoder now mixes
|
| 71 |
+
noise into half of the pulse power above 1 kHz (mixed excitation, no extra
|
| 72 |
+
parameters), which removes most of the earlier mechanical buzz; a faint
|
| 73 |
+
trace can remain. Use the current code with these weights: older code loads
|
| 74 |
+
them without the mix.
|
| 75 |
- eSpeak NG misses some liaisons (*comment‿allez*), adds a few false ones, and
|
| 76 |
reads *fils* (son) as [fil].
|
| 77 |
- Two fixed voices. No voice cloning.
|
config.json
CHANGED
|
@@ -33,5 +33,15 @@
|
|
| 33 |
"mel_scale": 5.0,
|
| 34 |
"energy_mean": -4.0,
|
| 35 |
"energy_scale": 2.0,
|
| 36 |
-
"pitch_reference_hz": 200.0
|
| 37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
"mel_scale": 5.0,
|
| 34 |
"energy_mean": -4.0,
|
| 35 |
"energy_scale": 2.0,
|
| 36 |
+
"pitch_reference_hz": 200.0,
|
| 37 |
+
"aperiodicity": [
|
| 38 |
+
[
|
| 39 |
+
0.0,
|
| 40 |
+
0.0
|
| 41 |
+
],
|
| 42 |
+
[
|
| 43 |
+
1000.0,
|
| 44 |
+
0.5
|
| 45 |
+
]
|
| 46 |
+
]
|
| 47 |
+
}
|
onnx/model.onnx
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:698249cf63768ecd6011d9ddc4137ed16b4d2856a3bc3ba1ab7e1e4e153fc727
|
| 3 |
+
size 21368019
|
samples/FR_M1-1.wav
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 258092
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a633eabccfc7b6563e0c17b6973f433d58badf8fd5615b69edf07af6361c6395
|
| 3 |
size 258092
|
samples/FR_M1-2.wav
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 206380
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02332d3d78a5f2432b9d7f007883925556dbb86da892a048cf93c8df793ab5fa
|
| 3 |
size 206380
|
samples/FR_M1-3.wav
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 216108
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ae679287c68d2feceb5f5175e25ddc12d42ee202c64917ad9352bed1dccd8952
|
| 3 |
size 216108
|
samples/FR_M1-4.wav
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 196140
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f76f58b85100b97bd665b4374bfd1e5abaa54f090efd4b341d7d5bd108525b9b
|
| 3 |
size 196140
|
tiny-voca-fr.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:61c00f50dffeda1cc8773243b4722a1d64b57a016db90ecd96477aae6225b3a9
|
| 3 |
+
size 18191813
|