mova-tts 0.1.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,10 @@
1
+ .venv/
2
+ models/
3
+ *.egg-info/
4
+ __pycache__/
5
+ .pytest_cache/
6
+ dist/
7
+ build/
8
+ *.wav
9
+ !examples/*.wav
10
+ *.npz
mova_tts-0.1.0/LICENSE ADDED
@@ -0,0 +1,201 @@
1
+ Apache License
2
+ Version 2.0, January 2004
3
+ http://www.apache.org/licenses/
4
+
5
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
6
+
7
+ 1. Definitions.
8
+
9
+ "License" shall mean the terms and conditions for use, reproduction,
10
+ and distribution as defined by Sections 1 through 9 of this document.
11
+
12
+ "Licensor" shall mean the copyright owner or entity authorized by
13
+ the copyright owner that is granting the License.
14
+
15
+ "Legal Entity" shall mean the union of the acting entity and all
16
+ other entities that control, are controlled by, or are under common
17
+ control with that entity. For the purposes of this definition,
18
+ "control" means (i) the power, direct or indirect, to cause the
19
+ direction or management of such entity, whether by contract or
20
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
21
+ outstanding shares, or (iii) beneficial ownership of such entity.
22
+
23
+ "You" (or "Your") shall mean an individual or Legal Entity
24
+ exercising permissions granted by this License.
25
+
26
+ "Source" form shall mean the preferred form for making modifications,
27
+ including but not limited to software source code, documentation
28
+ source, and configuration files.
29
+
30
+ "Object" form shall mean any form resulting from mechanical
31
+ transformation or translation of a Source form, including but
32
+ not limited to compiled object code, generated documentation,
33
+ and conversions to other media types.
34
+
35
+ "Work" shall mean the work of authorship, whether in Source or
36
+ Object form, made available under the License, as indicated by a
37
+ copyright notice that is included in or attached to the work
38
+ (an example is provided in the Appendix below).
39
+
40
+ "Derivative Works" shall mean any work, whether in Source or Object
41
+ form, that is based on (or derived from) the Work and for which the
42
+ editorial revisions, annotations, elaborations, or other modifications
43
+ represent, as a whole, an original work of authorship. For the purposes
44
+ of this License, Derivative Works shall not include works that remain
45
+ separable from, or merely link (or bind by name) to the interfaces of,
46
+ the Work and Derivative Works thereof.
47
+
48
+ "Contribution" shall mean any work of authorship, including
49
+ the original version of the Work and any modifications or additions
50
+ to that Work or Derivative Works thereof, that is intentionally
51
+ submitted to Licensor for inclusion in the Work by the copyright owner
52
+ or by an individual or Legal Entity authorized to submit on behalf of
53
+ the copyright owner. For the purposes of this definition, "submitted"
54
+ means any form of electronic, verbal, or written communication sent
55
+ to the Licensor or its representatives, including but not limited to
56
+ communication on electronic mailing lists, source code control systems,
57
+ and issue tracking systems that are managed by, or on behalf of, the
58
+ Licensor for the purpose of discussing and improving the Work, but
59
+ excluding communication that is conspicuously marked or otherwise
60
+ designated in writing by the copyright owner as "Not a Contribution."
61
+
62
+ "Contributor" shall mean Licensor and any individual or Legal Entity
63
+ on behalf of whom a Contribution has been received by Licensor and
64
+ subsequently incorporated within the Work.
65
+
66
+ 2. Grant of Copyright License. Subject to the terms and conditions of
67
+ this License, each Contributor hereby grants to You a perpetual,
68
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
69
+ copyright license to reproduce, prepare Derivative Works of,
70
+ publicly display, publicly perform, sublicense, and distribute the
71
+ Work and such Derivative Works in Source or Object form.
72
+
73
+ 3. Grant of Patent License. Subject to the terms and conditions of
74
+ this License, each Contributor hereby grants to You a perpetual,
75
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
76
+ (except as stated in this section) patent license to make, have made,
77
+ use, offer to sell, sell, import, and otherwise transfer the Work,
78
+ where such license applies only to those patent claims licensable
79
+ by such Contributor that are necessarily infringed by their
80
+ Contribution(s) alone or by combination of their Contribution(s)
81
+ with the Work to which such Contribution(s) was submitted. If You
82
+ institute patent litigation against any entity (including a
83
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
84
+ or a Contribution incorporated within the Work constitutes direct
85
+ or contributory patent infringement, then any patent licenses
86
+ granted to You under this License for that Work shall terminate
87
+ as of the date such litigation is filed.
88
+
89
+ 4. Redistribution. You may reproduce and distribute copies of the
90
+ Work or Derivative Works thereof in any medium, with or without
91
+ modifications, and in Source or Object form, provided that You
92
+ meet the following conditions:
93
+
94
+ (a) You must give any other recipients of the Work or
95
+ Derivative Works a copy of this License; and
96
+
97
+ (b) You must cause any modified files to carry prominent notices
98
+ stating that You changed the files; and
99
+
100
+ (c) You must retain, in the Source form of any Derivative Works
101
+ that You distribute, all copyright, patent, trademark, and
102
+ attribution notices from the Source form of the Work,
103
+ excluding those notices that do not pertain to any part of
104
+ the Derivative Works; and
105
+
106
+ (d) If the Work includes a "NOTICE" text file as part of its
107
+ distribution, then any Derivative Works that You distribute must
108
+ include a readable copy of the attribution notices contained
109
+ within such NOTICE file, excluding those notices that do not
110
+ pertain to any part of the Derivative Works, in at least one
111
+ of the following places: within a NOTICE text file distributed
112
+ as part of the Derivative Works; within the Source form or
113
+ documentation, if provided along with the Derivative Works; or,
114
+ within a display generated by the Derivative Works, if and
115
+ wherever such third-party notices normally appear. The contents
116
+ of the NOTICE file are for informational purposes only and
117
+ do not modify the License. You may add Your own attribution
118
+ notices within Derivative Works that You distribute, alongside
119
+ or as an addendum to the NOTICE text from the Work, provided
120
+ that such additional attribution notices cannot be construed
121
+ as modifying the License.
122
+
123
+ You may add Your own copyright statement to Your modifications and
124
+ may provide additional or different license terms and conditions
125
+ for use, reproduction, or distribution of Your modifications, or
126
+ for any such Derivative Works as a whole, provided Your use,
127
+ reproduction, and distribution of the Work otherwise complies with
128
+ the conditions stated in this License.
129
+
130
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
131
+ any Contribution intentionally submitted for inclusion in the Work
132
+ by You to the Licensor shall be under the terms and conditions of
133
+ this License, without any additional terms or conditions.
134
+ Notwithstanding the above, nothing herein shall supersede or modify
135
+ the terms of any separate license agreement you may have executed
136
+ with Licensor regarding such Contributions.
137
+
138
+ 6. Trademarks. This License does not grant permission to use the trade
139
+ names, trademarks, service marks, or product names of the Licensor,
140
+ except as required for reasonable and customary use in describing the
141
+ origin of the Work and reproducing the content of the NOTICE file.
142
+
143
+ 7. Disclaimer of Warranty. Unless required by applicable law or
144
+ agreed to in writing, Licensor provides the Work (and each
145
+ Contributor provides its Contributions) on an "AS IS" BASIS,
146
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
147
+ implied, including, without limitation, any warranties or conditions
148
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
149
+ PARTICULAR PURPOSE. You are solely responsible for determining the
150
+ appropriateness of using or redistributing the Work and assume any
151
+ risks associated with Your exercise of permissions under this License.
152
+
153
+ 8. Limitation of Liability. In no event and under no legal theory,
154
+ whether in tort (including negligence), contract, or otherwise,
155
+ unless required by applicable law (such as deliberate and grossly
156
+ negligent acts) or agreed to in writing, shall any Contributor be
157
+ liable to You for damages, including any direct, indirect, special,
158
+ incidental, or consequential damages of any character arising as a
159
+ result of this License or out of the use or inability to use the
160
+ Work (including but not limited to damages for loss of goodwill,
161
+ work stoppage, computer failure or malfunction, or any and all
162
+ other commercial damages or losses), even if such Contributor
163
+ has been advised of the possibility of such damages.
164
+
165
+ 9. Accepting Warranty or Additional Liability. While redistributing
166
+ the Work or Derivative Works thereof, You may choose to offer,
167
+ and charge a fee for, acceptance of support, warranty, indemnity,
168
+ or other liability obligations and/or rights consistent with this
169
+ License. However, in accepting such obligations, You may act only
170
+ on Your own behalf and on Your sole responsibility, not on behalf
171
+ of any other Contributor, and only if You agree to indemnify,
172
+ defend, and hold each Contributor harmless for any liability
173
+ incurred by, or claims asserted against, such Contributor by reason
174
+ of your accepting any such warranty or additional liability.
175
+
176
+ END OF TERMS AND CONDITIONS
177
+
178
+ APPENDIX: How to apply the Apache License to your work.
179
+
180
+ To apply the Apache License to your work, attach the following
181
+ boilerplate notice, with the fields enclosed by brackets "[]"
182
+ replaced with your own identifying information. (Don't include
183
+ the brackets!) The text should be enclosed in the appropriate
184
+ comment syntax for the file format. We also recommend that a
185
+ file or class name and description of purpose be included on the
186
+ same "printed page" as the copyright notice for easier
187
+ identification within third-party archives.
188
+
189
+ Copyright [yyyy] [name of copyright owner]
190
+
191
+ Licensed under the Apache License, Version 2.0 (the "License");
192
+ you may not use this file except in compliance with the License.
193
+ You may obtain a copy of the License at
194
+
195
+ http://www.apache.org/licenses/LICENSE-2.0
196
+
197
+ Unless required by applicable law or agreed to in writing, software
198
+ distributed under the License is distributed on an "AS IS" BASIS,
199
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
200
+ See the License for the specific language governing permissions and
201
+ limitations under the License.
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2022 lang-uk
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
@@ -0,0 +1,161 @@
1
+ Metadata-Version: 2.5
2
+ Name: mova-tts
3
+ Version: 0.1.0
4
+ Summary: Ukrainian text-to-speech with voice cloning: a NeuTTS-Air fine-tune, streaming on CPU, Apple and NVIDIA GPUs
5
+ Project-URL: Repository, https://github.com/MykhailoMS/mova-tts
6
+ Project-URL: Issues, https://github.com/MykhailoMS/mova-tts/issues
7
+ Project-URL: Model, https://huggingface.co/Mikhailo/mova-tts
8
+ Author: Mykhailo Mulesa
9
+ License-Expression: Apache-2.0
10
+ License-File: LICENSE
11
+ License-File: LICENSES/ipa-uk.txt
12
+ Keywords: gguf,neutts,speech,text-to-speech,tts,ukrainian,voice cloning
13
+ Classifier: Development Status :: 4 - Beta
14
+ Classifier: Intended Audience :: Developers
15
+ Classifier: Intended Audience :: Science/Research
16
+ Classifier: Natural Language :: Ukrainian
17
+ Classifier: Operating System :: OS Independent
18
+ Classifier: Programming Language :: Python :: 3
19
+ Classifier: Programming Language :: Python :: 3.10
20
+ Classifier: Programming Language :: Python :: 3.11
21
+ Classifier: Programming Language :: Python :: 3.12
22
+ Classifier: Programming Language :: Python :: 3.13
23
+ Classifier: Topic :: Multimedia :: Sound/Audio :: Speech
24
+ Requires-Python: <3.14,>=3.10
25
+ Requires-Dist: ctranslate2>=4.0
26
+ Requires-Dist: huggingface-hub>=0.23
27
+ Requires-Dist: llama-cpp-python>=0.3.16
28
+ Requires-Dist: naholos>=0.1.0
29
+ Requires-Dist: neucodec>=0.0.4
30
+ Requires-Dist: numpy>=1.24
31
+ Requires-Dist: onnxruntime>=1.18
32
+ Requires-Dist: resemble-perth>=1.0.1
33
+ Requires-Dist: scipy>=1.10
34
+ Requires-Dist: sentencepiece>=0.2
35
+ Requires-Dist: soundfile>=0.12
36
+ Requires-Dist: torchao<0.18
37
+ Provides-Extra: test
38
+ Requires-Dist: pytest>=8; extra == 'test'
39
+ Description-Content-Type: text/markdown
40
+
41
+ # mova-tts
42
+
43
+ Ukrainian text-to-speech with voice cloning. The model is a fine-tune of
44
+ [neuphonic/neutts-air](https://huggingface.co/neuphonic/neutts-air) on ~1780 hours of Ukrainian
45
+ speech. It runs through llama.cpp (GGUF) with the [NeuCodec](https://huggingface.co/neuphonic/neucodec)
46
+ audio codec. It streams, with the first audio after 0.16 s on an RTX 4090, and it runs faster
47
+ than real time on a CPU.
48
+
49
+ ```bash
50
+ pip install mova-tts
51
+ ```
52
+
53
+ ```python
54
+ from mova_tts import MovaTTS
55
+
56
+ tts = MovaTTS() # downloads the model on first use
57
+ voice = tts.voice("me.wav", "Точний текст, який звучить у записі.")
58
+ wav = tts.synthesize("Доброго дня! Сьогодні я розповім вам про нашу нову систему.", voice)
59
+ tts.save(wav, "out.wav")
60
+ ```
61
+
62
+ Clone any voice from a clean 3–12 second recording plus the exact text spoken in it. Encode a
63
+ voice once and reuse it with `voice.save("me.npz")` and `Voice.load("me.npz")`. There is an
64
+ example voice from Common Voice (CC0) in `examples/`:
65
+
66
+ ```bash
67
+ mova-tts "Доброго дня! Як ваші справи?" --ref examples/voice_cv.wav \
68
+ --ref-text "$(cat examples/voice_cv.txt)" -o out.wav
69
+ ```
70
+
71
+ ## Streaming
72
+
73
+ ```python
74
+ for chunk in tts.stream("Перше речення. Друге речення.", voice):
75
+ play(chunk) # float32, 24 kHz mono
76
+ ```
77
+
78
+ Audio comes in one-second chunks. The first chunk is 0.2 s long and needs 35 generated codes.
79
+ Long text is generated sentence by sentence.
80
+
81
+ ## Speed
82
+
83
+ Time to first audio (front-end included) and real-time factor (RTF: generation time divided
84
+ by audio length). Median over 6 sentences, measured on one machine:
85
+
86
+ | hardware | first audio | RTF |
87
+ |---|---:|---:|
88
+ | NVIDIA RTX 4090 (Q8_0 or Q4_K_M) | 0.16 s | 0.14 |
89
+ | Intel i9-13900KF, CPU only, 8 threads (Q8_0 or Q4_K_M) | 1.0 s | 0.72 |
90
+
91
+ `device="auto"` uses CUDA if it is available, then the Apple GPU (Metal), then the CPU. On the
92
+ CPU and on a Mac, the audio is decoded with the NeuCodec ONNX decoder. On the CPU, set
93
+ `threads=` to the number of performance cores. More threads than that slow it down.
94
+
95
+ ## Quality
96
+
97
+ The model was evaluated on 234 held-out sentences (Common Voice and audiobooks, speakers never
98
+ seen in training), cloning each speaker from a different sentence of theirs. Each sentence is
99
+ sampled, so every row gives the range over 3 seeds (2 for Q4_K_M):
100
+
101
+ | | WER | CER | UTMOS | speaker similarity |
102
+ |---|---:|---:|---:|---:|
103
+ | `synthesize()`, Q8_0 | 0.123–0.140 | 0.035–0.041 | 3.45 | 0.48–0.50 |
104
+ | `stream()`, Q8_0 | 0.135–0.140 | 0.041–0.046 | 3.37 | 0.48–0.50 |
105
+ | `stream()`, Q4_K_M | 0.131–0.135 | 0.041 | 3.33 | 0.48–0.49 |
106
+
107
+ WER and CER come from Whisper large-v3 transcripts of the generated audio. UTMOS is a
108
+ predicted naturalness score from 1 to 5. Speaker similarity is the ECAPA cosine to the
109
+ speaker's real recordings. A take that does not stop, or stops too early, is regenerated
110
+ with another seed. `stream()` can do that only before its first chunk is out.
111
+
112
+ ## Text front-end
113
+
114
+ 1. Clean-up, then numbers, dates and units turned into words with
115
+ [skypro1111/m2m100-ukr-verbalization](https://huggingface.co/skypro1111/m2m100-ukr-verbalization).
116
+ That model (~0.5 GB) downloads the first time a text contains digits.
117
+ 2. Latin words transliterated, with acronyms spelled out by letter names: CRM becomes сі-ар-ем.
118
+ 3. Stress from [naholos](https://github.com/MykhailoMS/naholos).
119
+ 4. IPA from [ipa-uk](https://github.com/patriotyk/ipa-uk), vendored here under MIT.
120
+
121
+ If a word is stressed wrong, mark it yourself and your mark always wins. You can write
122
+ `за+мок` or `за́мок` in the text, or keep a list of such words in a file:
123
+
124
+ ```python
125
+ tts = MovaTTS(user_dict="my_words.txt") # one stressed word per line: за+мок
126
+ print(tts.frontend.stressed("Старий замок")) # check the stress before synthesis
127
+ ```
128
+
129
+ ## Options
130
+
131
+ ```python
132
+ MovaTTS(quant="Q4_K_M") # smaller model: 0.6 GB instead of 0.8 GB
133
+ MovaTTS(device="cpu") # "auto" | "cuda" | "metal" | "cpu"
134
+ MovaTTS(seed=42) # repeatable takes; default is a new take every call
135
+ MovaTTS(watermark=False) # synthesize() adds a Perth watermark by default
136
+ ```
137
+
138
+ Each take is sampled, so it sounds a little different every time. Sometimes a take stretches
139
+ a word or pauses oddly. If that happens, generate again or use another seed.
140
+
141
+ ## Model and training
142
+
143
+ - Base: [neuphonic/neutts-air](https://huggingface.co/neuphonic/neutts-air), a 0.7B Qwen-based
144
+ speech language model (Apache-2.0). It was fully fine-tuned in two stages: 12k steps on a
145
+ 740-hour subset, then 20k steps on all the data.
146
+ - Data: ~1780 hours of Ukrainian speech: audiobooks read by professional narrators, podcasts,
147
+ Common Voice, and studio voice recordings. All audio was re-encoded with NeuCodec. Stress
148
+ labels come from the recordings themselves: the text was aligned with the Montreal Forced
149
+ Aligner, and an acoustic detector marked the stress each speaker actually used.
150
+ - Weights: [Mikhailo/mova-tts](https://huggingface.co/Mikhailo/mova-tts) holds safetensors (for
151
+ transformers) and GGUF Q8_0 and Q4_K_M files (for llama.cpp).
152
+
153
+ ## Responsible use
154
+
155
+ Clone only voices you have permission to use. `synthesize()` output carries the imperceptible
156
+ [Perth](https://github.com/resemble-ai/Perth) watermark, so generated audio can be detected.
157
+
158
+ ## License
159
+
160
+ Code and model: Apache-2.0, like NeuTTS-Air and NeuCodec. The vendored ipa-uk is MIT
161
+ (`LICENSES/ipa-uk.txt`).
@@ -0,0 +1,121 @@
1
+ # mova-tts
2
+
3
+ Ukrainian text-to-speech with voice cloning. The model is a fine-tune of
4
+ [neuphonic/neutts-air](https://huggingface.co/neuphonic/neutts-air) on ~1780 hours of Ukrainian
5
+ speech. It runs through llama.cpp (GGUF) with the [NeuCodec](https://huggingface.co/neuphonic/neucodec)
6
+ audio codec. It streams, with the first audio after 0.16 s on an RTX 4090, and it runs faster
7
+ than real time on a CPU.
8
+
9
+ ```bash
10
+ pip install mova-tts
11
+ ```
12
+
13
+ ```python
14
+ from mova_tts import MovaTTS
15
+
16
+ tts = MovaTTS() # downloads the model on first use
17
+ voice = tts.voice("me.wav", "Точний текст, який звучить у записі.")
18
+ wav = tts.synthesize("Доброго дня! Сьогодні я розповім вам про нашу нову систему.", voice)
19
+ tts.save(wav, "out.wav")
20
+ ```
21
+
22
+ Clone any voice from a clean 3–12 second recording plus the exact text spoken in it. Encode a
23
+ voice once and reuse it with `voice.save("me.npz")` and `Voice.load("me.npz")`. There is an
24
+ example voice from Common Voice (CC0) in `examples/`:
25
+
26
+ ```bash
27
+ mova-tts "Доброго дня! Як ваші справи?" --ref examples/voice_cv.wav \
28
+ --ref-text "$(cat examples/voice_cv.txt)" -o out.wav
29
+ ```
30
+
31
+ ## Streaming
32
+
33
+ ```python
34
+ for chunk in tts.stream("Перше речення. Друге речення.", voice):
35
+ play(chunk) # float32, 24 kHz mono
36
+ ```
37
+
38
+ Audio comes in one-second chunks. The first chunk is 0.2 s long and needs 35 generated codes.
39
+ Long text is generated sentence by sentence.
40
+
41
+ ## Speed
42
+
43
+ Time to first audio (front-end included) and real-time factor (RTF: generation time divided
44
+ by audio length). Median over 6 sentences, measured on one machine:
45
+
46
+ | hardware | first audio | RTF |
47
+ |---|---:|---:|
48
+ | NVIDIA RTX 4090 (Q8_0 or Q4_K_M) | 0.16 s | 0.14 |
49
+ | Intel i9-13900KF, CPU only, 8 threads (Q8_0 or Q4_K_M) | 1.0 s | 0.72 |
50
+
51
+ `device="auto"` uses CUDA if it is available, then the Apple GPU (Metal), then the CPU. On the
52
+ CPU and on a Mac, the audio is decoded with the NeuCodec ONNX decoder. On the CPU, set
53
+ `threads=` to the number of performance cores. More threads than that slow it down.
54
+
55
+ ## Quality
56
+
57
+ The model was evaluated on 234 held-out sentences (Common Voice and audiobooks, speakers never
58
+ seen in training), cloning each speaker from a different sentence of theirs. Each sentence is
59
+ sampled, so every row gives the range over 3 seeds (2 for Q4_K_M):
60
+
61
+ | | WER | CER | UTMOS | speaker similarity |
62
+ |---|---:|---:|---:|---:|
63
+ | `synthesize()`, Q8_0 | 0.123–0.140 | 0.035–0.041 | 3.45 | 0.48–0.50 |
64
+ | `stream()`, Q8_0 | 0.135–0.140 | 0.041–0.046 | 3.37 | 0.48–0.50 |
65
+ | `stream()`, Q4_K_M | 0.131–0.135 | 0.041 | 3.33 | 0.48–0.49 |
66
+
67
+ WER and CER come from Whisper large-v3 transcripts of the generated audio. UTMOS is a
68
+ predicted naturalness score from 1 to 5. Speaker similarity is the ECAPA cosine to the
69
+ speaker's real recordings. A take that does not stop, or stops too early, is regenerated
70
+ with another seed. `stream()` can do that only before its first chunk is out.
71
+
72
+ ## Text front-end
73
+
74
+ 1. Clean-up, then numbers, dates and units turned into words with
75
+ [skypro1111/m2m100-ukr-verbalization](https://huggingface.co/skypro1111/m2m100-ukr-verbalization).
76
+ That model (~0.5 GB) downloads the first time a text contains digits.
77
+ 2. Latin words transliterated, with acronyms spelled out by letter names: CRM becomes сі-ар-ем.
78
+ 3. Stress from [naholos](https://github.com/MykhailoMS/naholos).
79
+ 4. IPA from [ipa-uk](https://github.com/patriotyk/ipa-uk), vendored here under MIT.
80
+
81
+ If a word is stressed wrong, mark it yourself and your mark always wins. You can write
82
+ `за+мок` or `за́мок` in the text, or keep a list of such words in a file:
83
+
84
+ ```python
85
+ tts = MovaTTS(user_dict="my_words.txt") # one stressed word per line: за+мок
86
+ print(tts.frontend.stressed("Старий замок")) # check the stress before synthesis
87
+ ```
88
+
89
+ ## Options
90
+
91
+ ```python
92
+ MovaTTS(quant="Q4_K_M") # smaller model: 0.6 GB instead of 0.8 GB
93
+ MovaTTS(device="cpu") # "auto" | "cuda" | "metal" | "cpu"
94
+ MovaTTS(seed=42) # repeatable takes; default is a new take every call
95
+ MovaTTS(watermark=False) # synthesize() adds a Perth watermark by default
96
+ ```
97
+
98
+ Each take is sampled, so it sounds a little different every time. Sometimes a take stretches
99
+ a word or pauses oddly. If that happens, generate again or use another seed.
100
+
101
+ ## Model and training
102
+
103
+ - Base: [neuphonic/neutts-air](https://huggingface.co/neuphonic/neutts-air), a 0.7B Qwen-based
104
+ speech language model (Apache-2.0). It was fully fine-tuned in two stages: 12k steps on a
105
+ 740-hour subset, then 20k steps on all the data.
106
+ - Data: ~1780 hours of Ukrainian speech: audiobooks read by professional narrators, podcasts,
107
+ Common Voice, and studio voice recordings. All audio was re-encoded with NeuCodec. Stress
108
+ labels come from the recordings themselves: the text was aligned with the Montreal Forced
109
+ Aligner, and an acoustic detector marked the stress each speaker actually used.
110
+ - Weights: [Mikhailo/mova-tts](https://huggingface.co/Mikhailo/mova-tts) holds safetensors (for
111
+ transformers) and GGUF Q8_0 and Q4_K_M files (for llama.cpp).
112
+
113
+ ## Responsible use
114
+
115
+ Clone only voices you have permission to use. `synthesize()` output carries the imperceptible
116
+ [Perth](https://github.com/resemble-ai/Perth) watermark, so generated audio can be detected.
117
+
118
+ ## License
119
+
120
+ Code and model: Apache-2.0, like NeuTTS-Air and NeuCodec. The vendored ipa-uk is MIT
121
+ (`LICENSES/ipa-uk.txt`).
@@ -0,0 +1 @@
1
+ Він доклав великих зусиль, спрямованих на відродження та виживання кримського народу.
Binary file
@@ -0,0 +1,62 @@
1
+ [build-system]
2
+ requires = ["hatchling>=1.25"]
3
+ build-backend = "hatchling.build"
4
+
5
+ [project]
6
+ name = "mova-tts"
7
+ version = "0.1.0"
8
+ description = "Ukrainian text-to-speech with voice cloning: a NeuTTS-Air fine-tune, streaming on CPU, Apple and NVIDIA GPUs"
9
+ readme = "README.md"
10
+ license = "Apache-2.0"
11
+ license-files = ["LICENSE", "LICENSES/*"]
12
+ requires-python = ">=3.10,<3.14"
13
+ authors = [{ name = "Mykhailo Mulesa" }]
14
+ keywords = ["ukrainian", "tts", "text-to-speech", "voice cloning", "neutts", "gguf", "speech"]
15
+ classifiers = [
16
+ "Programming Language :: Python :: 3",
17
+ "Programming Language :: Python :: 3.10",
18
+ "Programming Language :: Python :: 3.11",
19
+ "Programming Language :: Python :: 3.12",
20
+ "Programming Language :: Python :: 3.13",
21
+ "Operating System :: OS Independent",
22
+ "Development Status :: 4 - Beta",
23
+ "Intended Audience :: Developers",
24
+ "Intended Audience :: Science/Research",
25
+ "Natural Language :: Ukrainian",
26
+ "Topic :: Multimedia :: Sound/Audio :: Speech",
27
+ ]
28
+ dependencies = [
29
+ "naholos>=0.1.0",
30
+ "llama-cpp-python>=0.3.16",
31
+ "neucodec>=0.0.4",
32
+ "torchao<0.18", # neucodec -> torchtune 0.6 imports torchao.dtypes.nf4tensor (gone in 0.18)
33
+ "onnxruntime>=1.18",
34
+ "resemble-perth>=1.0.1",
35
+ "soundfile>=0.12",
36
+ "scipy>=1.10",
37
+ "numpy>=1.24",
38
+ "huggingface-hub>=0.23",
39
+ "ctranslate2>=4.0",
40
+ "sentencepiece>=0.2",
41
+ ]
42
+
43
+ [project.optional-dependencies]
44
+ test = ["pytest>=8"]
45
+
46
+ [project.scripts]
47
+ mova-tts = "mova_tts.cli:main"
48
+
49
+ [project.urls]
50
+ Repository = "https://github.com/MykhailoMS/mova-tts"
51
+ Issues = "https://github.com/MykhailoMS/mova-tts/issues"
52
+ Model = "https://huggingface.co/Mikhailo/mova-tts"
53
+
54
+ [tool.hatch.build.targets.wheel]
55
+ packages = ["src/mova_tts"]
56
+
57
+ [tool.hatch.build.targets.sdist]
58
+ include = ["src/mova_tts", "tests", "examples", "README.md", "LICENSE", "LICENSES", "pyproject.toml"]
59
+
60
+ [tool.pytest.ini_options]
61
+ testpaths = ["tests"]
62
+ markers = ["slow: downloads the model (~0.8 GB) and synthesizes"]
@@ -0,0 +1,6 @@
1
+ """mova-tts - Ukrainian TTS with voice cloning (NeuTTS-Air fine-tune)."""
2
+ from .frontend import Frontend
3
+ from .tts import DEFAULT_MODEL, SR, MovaTTS, Voice
4
+
5
+ __version__ = "0.1.0"
6
+ __all__ = ["MovaTTS", "Voice", "Frontend", "DEFAULT_MODEL", "SR"]