README.md
9.1 KB · 785 lines · markdown Raw
1 ---
2 base_model:
3 - Qwen/Qwen3-0.6B
4 language:
5 - aae
6 - aal
7 - aao
8 - ab
9 - abb
10 - abn
11 - abr
12 - abs
13 - abv
14 - acm
15 - acw
16 - acx
17 - adf
18 - adx
19 - ady
20 - aeb
21 - aec
22 - af
23 - afb
24 - afo
25 - ahl
26 - ahs
27 - ajg
28 - aju
29 - ala
30 - aln
31 - alo
32 - am
33 - amu
34 - an
35 - anc
36 - ank
37 - anp
38 - anw
39 - aom
40 - apc
41 - apd
42 - arb
43 - arq
44 - ars
45 - ary
46 - arz
47 - as
48 - ast
49 - avl
50 - awo
51 - ayl
52 - ayp
53 - az
54 - ba
55 - bag
56 - bas
57 - bax
58 - bba
59 - bbj
60 - bbl
61 - bbu
62 - bce
63 - bci
64 - bcs
65 - bcy
66 - bda
67 - bde
68 - bdm
69 - be
70 - beb
71 - bew
72 - bfd
73 - bft
74 - bg
75 - bgp
76 - bhb
77 - bhh
78 - bho
79 - bhp
80 - bhr
81 - bjj
82 - bjk
83 - bjn
84 - bjt
85 - bkh
86 - bkm
87 - bky
88 - bmm
89 - bmq
90 - bn
91 - bnm
92 - bnn
93 - bns
94 - bo
95 - bou
96 - bqg
97 - br
98 - bra
99 - brh
100 - bri
101 - brx
102 - bs
103 - bsh
104 - bsj
105 - bsk
106 - btm
107 - btv
108 - bug
109 - bum
110 - buo
111 - bux
112 - bwr
113 - bxf
114 - byc
115 - bys
116 - byv
117 - byx
118 - bzc
119 - bzw
120 - ca
121 - ccg
122 - ceb
123 - cen
124 - cfa
125 - cgg
126 - chq
127 - cjk
128 - ckb
129 - ckl
130 - ckr
131 - cky
132 - cnh
133 - cpy
134 - cs
135 - cte
136 - ctl
137 - cut
138 - cux
139 - cv
140 - cy
141 - da
142 - dag
143 - dar
144 - dav
145 - dbd
146 - dcc
147 - de
148 - deg
149 - dgh
150 - dgo
151 - dje
152 - dmk
153 - dml
154 - dru
155 - dty
156 - dua
157 - dv
158 - dyu
159 - dzg
160 - ebr
161 - ebu
162 - ego
163 - eiv
164 - eko
165 - ekr
166 - el
167 - elm
168 - en
169 - eo
170 - es
171 - esu
172 - et
173 - eto
174 - ets
175 - etu
176 - eu
177 - ewo
178 - ext
179 - eyo
180 - fa
181 - fan
182 - fat
183 - ff
184 - ffm
185 - fi
186 - fia
187 - fil
188 - fip
189 - fkk
190 - fmp
191 - fr
192 - fub
193 - fuc
194 - fue
195 - fuf
196 - fuh
197 - fui
198 - fuq
199 - fuv
200 - fy
201 - ga
202 - gbm
203 - gbr
204 - gby
205 - gcc
206 - gdf
207 - gej
208 - ges
209 - ggg
210 - gid
211 - gig
212 - giz
213 - gjk
214 - gju
215 - gl
216 - glw
217 - gn
218 - gol
219 - gom
220 - gsl
221 - gu
222 - gui
223 - gur
224 - guz
225 - gv
226 - gwc
227 - gwe
228 - gwt
229 - gya
230 - gyz
231 - ha
232 - hah
233 - hao
234 - haw
235 - haz
236 - hbb
237 - he
238 - hem
239 - hi
240 - hia
241 - hkk
242 - hla
243 - hno
244 - hoj
245 - hr
246 - hsb
247 - ht
248 - hu
249 - hue
250 - hul
251 - hux
252 - hwo
253 - hy
254 - hz
255 - ia
256 - ibb
257 - id
258 - ida
259 - idu
260 - ig
261 - ijc
262 - ijn
263 - ik
264 - ikw
265 - is
266 - ish
267 - iso
268 - it
269 - its
270 - itw
271 - itz
272 - ja
273 - jal
274 - jax
275 - jgo
276 - jmx
277 - jns
278 - jqr
279 - juk
280 - juo
281 - jv
282 - ka
283 - kab
284 - kai
285 - kaj
286 - kam
287 - kbd
288 - kbl
289 - kbt
290 - kcq
291 - kdh
292 - kea
293 - keu
294 - kfe
295 - kfk
296 - kfp
297 - khg
298 - khw
299 - kj
300 - kjc
301 - kjk
302 - kk
303 - kln
304 - kls
305 - km
306 - kmr
307 - kmy
308 - kn
309 - kna
310 - knn
311 - ko
312 - kol
313 - koo
314 - kpo
315 - kqo
316 - ks
317 - ksd
318 - ksf
319 - kto
320 - kuh
321 - kvx
322 - kw
323 - kwm
324 - kxp
325 - ky
326 - kyx
327 - lag
328 - lb
329 - lcm
330 - ldb
331 - lg
332 - lij
333 - lir
334 - lkb
335 - lla
336 - ln
337 - lnu
338 - lo
339 - loa
340 - lrk
341 - lss
342 - lt
343 - ltg
344 - lto
345 - lua
346 - luo
347 - lus
348 - lv
349 - lwg
350 - mab
351 - maf
352 - mai
353 - mau
354 - max
355 - mbo
356 - mcf
357 - mcn
358 - mcx
359 - mdd
360 - mde
361 - mdf
362 - mek
363 - mer
364 - meu
365 - mfm
366 - mfn
367 - mfo
368 - mfv
369 - mgg
370 - mgi
371 - mhk
372 - mhr
373 - mi
374 - mig
375 - miu
376 - mk
377 - mkf
378 - mki
379 - ml
380 - mlq
381 - mn
382 - mne
383 - mni
384 - mqy
385 - mr
386 - mrj
387 - mrr
388 - mrt
389 - ms
390 - mse
391 - msh
392 - msw
393 - mt
394 - mtr
395 - mtu
396 - mtx
397 - mua
398 - mug
399 - mui
400 - mve
401 - mvy
402 - mxs
403 - mxu
404 - mxy
405 - my
406 - myv
407 - mzl
408 - nal
409 - nan
410 - nap
411 - nb
412 - nbh
413 - ncf
414 - nco
415 - ncx
416 - ndi
417 - ng
418 - ngi
419 - nhg
420 - nhi
421 - nhn
422 - nhq
423 - nja
424 - nl
425 - nla
426 - nlv
427 - nmg
428 - nmz
429 - nn
430 - nnh
431 - 'no'
432 - noe
433 - npi
434 - nso
435 - ny
436 - nyu
437 - oc
438 - odk
439 - odu
440 - ogo
441 - om
442 - orc
443 - oru
444 - ory
445 - os
446 - pa
447 - pbs
448 - pbt
449 - pbu
450 - pcm
451 - pex
452 - phl
453 - phr
454 - pip
455 - piy
456 - pko
457 - pl
458 - plk
459 - plt
460 - pmq
461 - pms
462 - pmy
463 - pnb
464 - poc
465 - poe
466 - pow
467 - prq
468 - ps
469 - pst
470 - pt
471 - pua
472 - pwn
473 - qug
474 - qum
475 - qup
476 - qur
477 - qus
478 - quv
479 - qux
480 - quy
481 - qva
482 - qvi
483 - qvj
484 - qvl
485 - qwa
486 - qws
487 - qxa
488 - qxp
489 - qxt
490 - qxu
491 - qxw
492 - rag
493 - rm
494 - ro
495 - rob
496 - rof
497 - roo
498 - rth
499 - ru
500 - rup
501 - rw
502 - sa
503 - sah
504 - sat
505 - sau
506 - say
507 - sbn
508 - sc
509 - scl
510 - scn
511 - sd
512 - sei
513 - shu
514 - si
515 - sip
516 - siw
517 - sjr
518 - sk
519 - skg
520 - skr
521 - sl
522 - sn
523 - snc
524 - snk
525 - so
526 - sol
527 - sps
528 - sq
529 - sr
530 - src
531 - sro
532 - ssi
533 - ste
534 - sua
535 - sv
536 - sva
537 - sw
538 - szy
539 - ta
540 - tan
541 - tar
542 - tay
543 - tbf
544 - tcf
545 - tcy
546 - tdn
547 - tdx
548 - te
549 - tg
550 - tgc
551 - th
552 - the
553 - thq
554 - thr
555 - thv
556 - ti
557 - tig
558 - tio
559 - tk
560 - tkg
561 - tkt
562 - tli
563 - tlp
564 - tn
565 - tok
566 - tpl
567 - tpz
568 - tqp
569 - tr
570 - trp
571 - trq
572 - trv
573 - trw
574 - tt
575 - ttj
576 - ttr
577 - ttu
578 - tui
579 - tul
580 - tuq
581 - tuv
582 - tuy
583 - tvo
584 - tvu
585 - tw
586 - twu
587 - txs
588 - txy
589 - udl
590 - ug
591 - uk
592 - uki
593 - umb
594 - ur
595 - ush
596 - uz
597 - uzn
598 - vai
599 - var
600 - ver
601 - vi
602 - vmc
603 - vmj
604 - vmm
605 - vmp
606 - vmz
607 - vot
608 - vro
609 - wbl
610 - wci
611 - weo
612 - wes
613 - wja
614 - wji
615 - wo
616 - wof
617 - xh
618 - xhe
619 - xka
620 - xmf
621 - xmv
622 - xmw
623 - xpe
624 - xti
625 - xtu
626 - yaq
627 - yav
628 - yay
629 - ydd
630 - ydg
631 - yer
632 - 'yes'
633 - yi
634 - yo
635 - yue
636 - zga
637 - zgh
638 - zh
639 - zoc
640 - zoh
641 - zor
642 - zpv
643 - zpy
644 - ztg
645 - ztn
646 - ztp
647 - zts
648 - ztu
649 - zu
650 - zza
651 license: apache-2.0
652 pipeline_tag: text-to-speech
653 tags:
654 - zero-shot
655 - multilingual
656 - voice-cloning
657 - voice-design
658 library_name: omnivoice
659 ---
660
661 # OmniVoice 🌍
662
663 <p align="center">
664 <img width="200" height="200" alt="OmniVoice" src="https://zhu-han.github.io/omnivoice/pics/omnivoice.jpg" />
665 </p>
666
667 <p align="center">
668 <a href="https://huggingface.co/k2-fsa/OmniVoice"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Model-FFD21E" alt="Hugging Face Model"></a>
669 &nbsp;
670 <a href="https://huggingface.co/spaces/k2-fsa/OmniVoice"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Space-blue" alt="Hugging Face Space"></a>
671 &nbsp;
672 <a href="https://huggingface.co/papers/2604.00688"><img src="https://img.shields.io/badge/arXiv-Paper-B31B1B.svg"></a>
673 &nbsp;
674 <a href="https://github.com/k2-fsa/OmniVoice"><img src="https://img.shields.io/badge/GitHub-Code-181717?logo=GitHub" alt="GitHub Code"></a>
675 &nbsp;
676 <a href="https://zhu-han.github.io/omnivoice"><img src="https://img.shields.io/badge/GitHub.io-Demo_Page-blue?logo=GitHub&style=flat-square"></a>
677 &nbsp;
678 <a href="https://colab.research.google.com/github/k2-fsa/OmniVoice/blob/master/docs/OmniVoice.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>
679 </p>
680
681
682 OmniVoice is a massively multilingual zero-shot text-to-speech (TTS) model supporting over 600 languages. Built on a novel diffusion language model-style architecture, it delivers high-quality speech with superior inference speed, supporting voice cloning and voice design.
683
684 - **Paper:** [OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Models](https://huggingface.co/papers/2604.00688)
685 - **Repository:** [GitHub](https://github.com/k2-fsa/OmniVoice)
686 - **Demo:** [Hugging Face Space](https://huggingface.co/spaces/k2-fsa/OmniVoice)
687 - **Colab:** [Google Colab Notebook](https://colab.research.google.com/github/k2-fsa/OmniVoice/blob/master/docs/OmniVoice.ipynb)
688
689 ## Key Features
690
691 - **600+ Languages Supported**: The broadest language coverage among zero-shot TTS models.
692 - **Voice Cloning**: State-of-the-art voice cloning quality from a short reference audio.
693 - **Voice Design**: Control voices via assigned speaker attributes (gender, age, pitch, dialect/accent, whisper, etc.).
694 - **Fine-grained Control**: Non-verbal symbols (e.g., `[laughter]`) and pronunciation correction via pinyin or phonemes.
695 - **Fast Inference**: RTF as low as 0.025 (40x faster than real-time).
696 - **Diffusion Language Model-style Architecture**: A clean, streamlined, and scalable design that delivers both quality and speed.
697
698 ## Usage
699
700 To get started, install the `omnivoice` library:
701
702 > We recommend using a fresh virtual environment (e.g., `conda`, `venv`, etc.) to avoid conflicts.
703
704 **Step 1**: Install PyTorch
705
706 <details>
707 <summary>NVIDIA GPU</summary>
708
709 ```bash
710 # Install pytorch with your CUDA version, e.g.
711 pip install torch==2.8.0+cu128 torchaudio==2.8.0+cu128 --extra-index-url https://download.pytorch.org/whl/cu128
712 ```
713 > See [PyTorch official site](https://pytorch.org/get-started/locally/) for other versions installation.
714
715 </details>
716
717 <details>
718 <summary>Apple Silicon</summary>
719
720 ```bash
721 pip install torch==2.8.0 torchaudio==2.8.0
722 ```
723
724 </details>
725
726 **Step 2**: Install OmniVoice
727
728 ```bash
729 pip install omnivoice
730 ```
731
732 ### Python API
733
734 You can use OmniVoice for zero-shot voice cloning as follows:
735
736 ```python
737 from omnivoice import OmniVoice
738 import soundfile as sf
739 import torch
740
741 # Load the model
742 model = OmniVoice.from_pretrained(
743 "k2-fsa/OmniVoice",
744 device_map="cuda:0",
745 dtype=torch.float16
746 )
747
748 # Generate audio
749 audio = model.generate(
750 text="Hello, this is a test of zero-shot voice cloning.",
751 ref_audio="ref.wav",
752 ref_text="Transcription of the reference audio.",
753 ) # audio is a list of `np.ndarray` with shape (T,) at 24 kHz.
754
755 sf.write("out.wav", audio[0], 24000)
756 ```
757
758 For more generation modes (e.g., voice design), functions (e.g., non-verbal symbols, pronunciation correction) and comprehensive usage instructions, see our [GitHub Repository](https://github.com/k2-fsa/OmniVoice).
759
760
761 ## Discussion & Communication
762
763 You can directly discuss on [GitHub Issues](https://github.com/k2-fsa/OmniVoice/issues).
764
765 You can also scan the QR code to join our wechat group or follow our wechat official account.
766
767 | Wechat Group | Wechat Official Account |
768 | ------------ | ----------------------- |
769 |![wechat](https://k2-fsa.org/zh-CN/assets/pic/wechat_group.jpg) |![wechat](https://k2-fsa.org/zh-CN/assets/pic/wechat_account.jpg) |
770
771 ## Citation
772
773 ```bibtex
774 @article{zhu2026omnivoice,
775 title={OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Models},
776 author={Zhu, Han and Ye, Lingxuan and Kang, Wei and Yao, Zengwei and Guo, Liyong and Kuang, Fangjun and Han, Zhifeng and Zhuang, Weiji and Lin, Long and Povey, Daniel},
777 journal={arXiv preprint arXiv:2604.00688},
778 year={2026}
779 }
780 ```
781
782
783 ## Disclaimer
784
785 Users are strictly prohibited from using this model for unauthorized voice cloning, voice impersonation, fraud, scams, or any other illegal or unethical activities. All users shall ensure full compliance with applicable local laws, regulations, and ethical standards. The developers assume no liability for any misuse of this model and advocate for responsible AI development and use, encouraging the community to uphold safety and ethical principles in AI research and applications.