# VoiceStudio passes 85,000 stars: the local ElevenLabs alternative in numbers

> VoiceStudio gained 10,609 GitHub stars in the last week and 37,473 in 27 days. Here is what the activity data on RepoQuarry shows, who builds it, and what you need to know about the licence before you use it.

By Wayne Hewitt, 25 Oct 2026. Source: https://repoquarry.dev/blog/voicestudio-85000-stars-in-numbers

**VoiceStudio is a free desktop app that clones voices, dubs video, transcribes speech and builds audiobooks on your own computer. On 25 October 2026 it had 85,019 GitHub stars, up 10,609 in seven days. Six months ago the repository did not exist.**

It is the third most trending project on RepoQuarry this week, and the [VoiceStudio page](https://repoquarry.dev/projects/voicestudio) rates its growth as healthy. These are the numbers behind that label.

## Stars: 37,473 in 27 days

| Measure | Figure |
|---|---|
| GitHub stars | 85,019 |
| Gained in the last 7 days | +10,609 (+14.3%) |
| Gained in the last 27 days | +37,473 |
| Forks | 9,412 |
| Repository created | 9 April 2026 |

By the afternoon of the 25th GitHub was already showing more than 86,000.

![Line chart of VoiceStudio GitHub stars from April to October 2026. The line stays under 700 until late August, then climbs to 54,181 by 5 October and 85,019 by 25 October.](https://repoquarry.dev/blog/voicestudio-85000-stars-in-numbers/stars-over-time.webp)

For its first four months VoiceStudio was a quiet project. On 24 August it had 635 stars. The climb started that week, alongside the v0.5.1 release on 28 August, and it has not slowed since.

![Bar chart of new GitHub stars per day from 16 to 24 October: 1,109, 2,058, 2,622, 1,639, 1,821, 1,151, 1,446, 1,329 and 2,460.](https://repoquarry.dev/blog/voicestudio-85000-stars-in-numbers/stars-per-day.webp)

Every day we have tracked, at least 1,100 people starred it. The weekend days were the biggest, with 2,622 on Sunday 18 October and 2,460 on Saturday 24 October.

The chart on the project page carries a note: "not enough baseline history". RepoQuarry started recording VoiceStudio every day on 15 October. The dashed part of the line before that date is estimated from GitHub's public record of when each person starred the repo. Ten days of our own readings are too few to compare this week against a normal week for the project, so we hold back a "faster than usual" or "slowing" label until we have more. The 7-day and 27-day gains above are measured, not estimated.

## Commits: 12 busy weeks in a row

The commit chart on our page covers the last 52 weeks. The first 22 are empty, because the project did not exist yet. Since the first commit in early April:

- commits landed in every single week
- the last 12 weeks had 2,459 commits, against 927 in the 12 weeks before
- the busiest week, starting 27 September, had 337 commits
- GitHub counts 5,156 commits on the main branch in total

![Bar chart of VoiceStudio commits per week from April to October 2026. Weekly commits stay under 160 until August, then reach 330, 305 and a peak of 337 in September, and 324 in the week of 18 October.](https://repoquarry.dev/blog/voicestudio-85000-stars-in-numbers/commits-per-week.webp)

The work picked up before the stars did. The week of 9 August had 253 commits, three weeks ahead of the jump in stars.

Other people are doing more than starring it. In the last 30 days the project merged 268 pull requests, and 279 issues were opened against 274 closed. More than 100 people have committed code.

Most of the code still comes from two people. Palash Debnath has 3,722 commits. Shimul Debnath ([velixio](https://github.com/velixio)), listed with him as core team in the README, has 723. The next contributor has 58.

## Releases

The pace shows in the release notes too:

- **v0.5.3, 17 September:** a new look, and a move from Tauri to Electron for the desktop app
- **v0.5.4, 21 September:** working engines and smoother long-form audio
- **v0.5.5, 22 September:** cross-language voice cloning
- **v0.5.6, 23 September:** the current stable version
- **v0.5.7-beta.2, 23 October:** a desktop beta, followed on 24 October by a mobile beta

Installers cover Windows, macOS on Apple Silicon and Linux, plus a Docker image. On Windows the one-line install is `irm https://voicestudio.sh/install | iex` in PowerShell. NVIDIA and Apple Silicon graphics speed things up. A processor alone works, slower, with about 5 GB of free disk. Intel Macs run the interface only and need a backend on another machine.

## Who builds it

VoiceStudio is the work of [Palash Debnath](https://palash.dev), a software engineer in Agartala, in northeast India. He has been on GitHub since 2013 and his profile bio lists him as CTO of [Yupcha](https://yupcha.com). The project started under the name OmniVoice Studio, after the [k2-fsa OmniVoice](https://huggingface.co/k2-fsa/OmniVoice) model that is still its default engine, and the old name survives in the repo's topics.

In July he wrote [Why I Built an Open-Source ElevenLabs](https://palash.dev/blog/omnivoice-open-source-elevenlabs/). His case: a cloud voice service holds your voice on someone else's servers, charges by the second, and keeps the right to change its terms. He treats a voice as part of your identity and wants you to own the whole stack. When he wrote the post the repo had nearly 10,000 stars. It has grown more than eightfold since.

VoiceStudio is one of several apps he ships. His site also lists memxt (local-first memory for AI agents), Opal (a media player), Bootable, and Friday, a local voice assistant for Linux he is still building.

Follow him and the project here:

- Website: [palash.dev](https://palash.dev)
- X: [@idebpalash](https://x.com/idebpalash)
- GitHub: [debpalash](https://github.com/debpalash)
- Project site: [voicestudio.sh](https://voicestudio.sh)
- Code: [github.com/debpalash/VoiceStudio](https://github.com/debpalash/VoiceStudio)
- Support the work: [Ko-fi](https://ko-fi.com/debpalash) or GitHub Sponsors

## Check the licence before you use it for work

For personal use, VoiceStudio replaces most of what you would pay ElevenLabs for: speech generation, voice cloning, transcription and dubbing. ElevenLabs paid plans start at $6 a month. VoiceStudio has no voice agents, which ElevenLabs sells.

Business use is different. The app code is AGPL-3.0, which allows commercial use if you share your changes. The default OmniVoice model weights are CC-BY-NC, which does not. So the setup you get out of the box is for non-commercial use only. Other engines in the app carry their own terms, and the project sells a commercial licence and Pro features if you need them. Our [facts table](https://repoquarry.dev/projects/voicestudio) links each of these to its source.

If you need voice cloning you can sell, [GPT-SoVITS](https://repoquarry.dev/projects/gpt-sovits) is MIT-licensed for both code and weights. For transcription alone, [WhisperX](https://repoquarry.dev/projects/whisperx) adds word timings and speaker labels to Whisper.

We will update this post when the star chart has enough history for a proper trend. Follow VoiceStudio on RepoQuarry and we email you if its licence changes.

## Projects in this post

- [VoiceStudio](https://repoquarry.dev/projects/voicestudio): VoiceStudio is an Electron desktop app (Windows, macOS Apple Silicon, Linux, Docker) with a Python backend that clones and designs voices, dubs video, transcribes and builds audiobooks on your own hardware; CPU works, slower. App code is AGPL-3.0, but the default engine's OmniVoice weights are CC-BY-NC, so commercial use of the default setup is not allowed. Other engines carry their own terms. A paid Pro tier and commercial licence exist. (https://repoquarry.dev/projects/voicestudio.md)
- [GPT-SoVITS](https://repoquarry.dev/projects/gpt-sovits): WebUI for zero-shot TTS from a 5-second sample or fine-tuning on 1 minute of audio, with cross-lingual output in English, Japanese, Korean, Cantonese and Chinese. Bundles dataset tools (vocal separation, slicing, ASR for labelling). MIT code and MIT pretrained weights on Hugging Face. Runs on NVIDIA, AMD (ROCm on Linux), Apple silicon and CPU; stated RTF 0.028 on an RTX 4060 Ti and 0.526 on an M4 CPU. Windows has a download-and-double-click package. (https://repoquarry.dev/projects/gpt-sovits.md)
- [Fish Speech](https://repoquarry.dev/projects/fish-speech): Inference code and weights for Fish Audio Drama 3 (35B-A3B MoE) and S2-Pro (4B): multi-speaker TTS, voice cloning from 10-30 s of audio, inline emotion tags. Not open source: code and Drama 3 use the Fish Audio Community License (commercial use only under USD 500k annual revenue, no hosted TTS services); S2-Pro weights are research and non-commercial only. Heavy hardware: Drama 3 needs at least 48 GB VRAM on NVIDIA; S2-Pro 24 GB or CPU. Licence updated 13 Oct 2026. (https://repoquarry.dev/projects/fish-speech.md)
- [WhisperX](https://repoquarry.dev/projects/whisperx): WhisperX wraps a faster-whisper backend with batched inference (README: 70x realtime with large-v2), wav2vec2 forced alignment for word timestamps, and pyannote speaker diarization. Code is BSD-2-Clause. Diarization uses the gated pyannote speaker-diarization-community-1 model (CC-BY-4.0) and needs a free Hugging Face token. README states under 8 GB GPU memory for large-v2; CPU mode is supported. (https://repoquarry.dev/projects/whisperx.md)