VoxPocket v3: small Khmer text-to-speech
VoxPocket is a 0.6B Khmer TTS model (44.1 kHz). It was fine-tuned from VoxCPM 1.5 on real Khmer speech and on synthetic speech distilled from VoxCPM2. It reads Khmer, Khmer mixed with English words, and plain English. It can clone a voice from a short clip and take a style or a voice description.
๐ฃ๏ธ Try it in your browser, no download needed: VoxPocket v3 demo. It has ready voices, styles and voice design.
Use
from voxcpm.core import VoxCPM
import soundfile as sf
m = VoxCPM.from_pretrained(hf_model_id="INGMUYLEANG/VoxPocket-v3", load_denoiser=False)
sr = m.tts_model.sample_rate # 44100
# 1) clone a voice: a clean 4โ9 s clip + exactly what is said in it
wav = m.generate(text="แแฝแแแแธแแแแแขแผแแแถแแแขแแแแแแถ แขแแแปแแ
แแแพแแแแแแถแแแแถแแแถแแแแแ",
prompt_wav_path="my_voice.wav", prompt_text="<words spoken in my_voice.wav>",
cfg_value=2.0, inference_timesteps=10)
sf.write("clone.wav", wav, sr)
# 2) clone + style: put the style in brackets before the text
wav = m.generate(text="(sad, low and slow tone)แแฝแแแแธแแแแแขแผแแแถแแแขแแแแแแถแ",
prompt_wav_path="my_voice.wav", prompt_text="<words spoken in my_voice.wav>")
# 3) design a voice, no clip: describe it in brackets
wav = m.generate(text="(A young woman, clear and bright voice, cheerful tone)แแฝแแแแธแแแแแขแผแแแถแแแขแแแแแแถแ")
Tips:
- Reference clips: keep them short, about 4โ9 s, and clean, with the exact words that are spoken.
- Long text: split it into sentences.
- Styles: the phrases used in training work best, for example
sad, low and slow tone,excited and energetic,whispering quietly,happy, smiling tone,calm, slow and steady,serious and clear news-reader tone.
Results
The test reads 40 held-out Khmer sentences with the same voice clip. A Khmer speech recognizer listens back, and we count letter errors (lower is better):
| model | median | mean | sentences with > 10% errors |
|---|---|---|---|
| VoxPocket v2 | 12.2% | 13.4% | 21 / 40 |
| VoxPocket v3 | 7.6% | 8.4% | 13 / 40 |
v3 is better on 24 sentences, the same on 6, and worse on 10.
Training
- Starting point: VoxPocket v2 (VoxCPM 1.5 base), fine-tuned on NVIDIA GPUs.
- Data: about 433 h in total.
- 237 h of real Khmer speech
- 177 h of synthetic VoxCPM2 speech, all checked by a speech recognizer: Khmer, mixed, English, clone, style and designed voices
- 18 h of clip + style pairs, so a cloned voice can change tone
- The synthetic part is published as INGMUYLEANG/voxpocket-khmer-tts-v3.
Limits
- Styles on cloned voices are still weak: the tone changes only a little. Designed voices (no clip) follow descriptions better.
- Rare words and names are sometimes swapped for common ones. English loanwords inside Khmer can sound Khmer-ish.
- English is basic. The model is made for Khmer first.
- Do not use it to impersonate real people or to make deceptive audio. Only clone voices you have permission to use.
- Downloads last month
- 6