ElevenLabs Alternatives on GitHub: Judge the Repository, Not the Star Count
By the SoftSea editorial team
Photo: Godfrey Atima / Pexels
A repository is evidence, and most of it is not the README
The GitHub-shaped version of this query attracts a different searcher: someone who will self-host, read code, and build on top. Incumbent roundups serve them badly, ranking projects by stars, which measure a moment of attention, not maintenance, legality, or quality. This site ran the same playbook for Airtable alternatives on GitHub, and the method transfers: judge the LICENSE file in the repo root, the license of the model weights (often a different document, sometimes a different universe), the date and cadence of releases, the issue queue's vital signs, and the archived banner that ends arguments. The open-source companion guide grades these projects by license alone; this page grades the repositories as things you might depend on.
Repository
What it is
Code license
Weights or model license
Maintenance signal
resemble-ai/chatterbox
Zero-shot cloning model
MIT
MIT release; PerTh watermark built in
Active, vendor-backed
hexgrad/kokoro
82M narration model
Apache-2.0
Apache-2.0
Active
SWivid/F5-TTS
Cloning-capable TTS
MIT
CC-BY-NC (non-commercial)
Active research
coqui-ai/TTS (XTTS v2)
Cloning model, 17 languages
MPL-2.0 toolkit
Coqui Public Model License, non-commercial
Frozen: company shut down
OHF-Voice/piper1-gpl
Offline TTS engine
GPL-3.0
Voices vary
Active; original rhasspy/piper archived Oct 6, 2025
VoiceStudio
Desktop TTS app
AGPL-3.0
Default engine OmniVoice is CC-BY-NC
Active app, engine trap inside
Qwen3-TTS
Alibaba model family
Check the repo LICENSE
Check the model card
New and moving fast
The five checks, in order
First, open the LICENSE file, not the README's license badge: badges lie by omission, and the file is the legal object. Second, find the weights license. Modern TTS is two artifacts (code and trained weights), and the split is where commercial dreams die: F5-TTS is MIT code over CC-BY-NC weights, so the demo is free and the product is not, unless you train or license weights separately. Third, read the releases page: a project with tagged releases in the last quarter is alive; a project whose last release predates the current model generation is a museum, however many stars it holds. Fourth, check for an archived banner or a pinned issue redirecting development, the Piper pattern: the original repository was archived on October 6, 2025, and the living project is piper1-gpl under the Open Home Foundation, a fact invisible from search snippets that still rank the dead repo. Fifth, skim the issues for the words "license", "commercial", and "weights": the maintainers' own answers there are the cheapest legal research you will ever do.
Interactive tool
Comparison Builder
Pick two or three products and see them side by side on the numbers that decide the purchase.
Measure
Chatterbox
Kokoro-82M
ElevenLabs
Starting paid price
Free
Free
$5 per month, flat
Free option
Free plan: fully free and open source (MIT)
Free plan: fully free and open source (Apache-2.0)
Free plan: 10,000 credits per month (about 10 minutes of speech), no commercial license
voiceover at production quality, API-driven narration, dubbing and localization
Strengths
Real zero-shot cloning, self-hosted; MIT license allows commercial use; Runs locally; audio never leaves your machine
Runs on a laptop CPU in real time; Apache-2.0: commercial use is clean; Quality punches far above its size
Most realistic generated voices in the category; Cloning quality sets the standard rivals chase; Credits work across text to speech, dubbing, and effects
Watch out for
You supply the GPU and the setup time; Watermarking is part of the design; No hosted support or SLA
No voice cloning at all; Voice count and languages are fixed; Assembly into a product is on you
Credits burn on every generation attempt, not just keepers; Credits never roll over; Free plan output cannot be used commercially
Lowest entry price: Chatterbox at Free per month flat. Real free plan: Chatterbox, Kokoro-82M, ElevenLabs. Entry price is the floor, not the decision: check the plan line above for the tier that actually includes the feature you need.
All figures from vendor pricing pages, checked November 18, 2026. Some vendors price by contacts, tasks, or operations instead of seats; those notes are in the plan data and the linked pricing guides.
The trap of the batch: VoiceStudio and OmniVoice
VoiceStudio deserves its own section because it is exactly what this query hopes to find (a friendly desktop app, open source, AGPL-3.0, cloning and narration without a subscription) and it contains the category's cleanest trap: the application is AGPL-3.0, but its default speech engine, OmniVoice, is licensed CC-BY-NC, non-commercial. The app license governs the app. The engine license governs your audio. A user can lawfully run VoiceStudio all day and still owe their commercial project a different engine. This is not a criticism of the project, which discloses its stack; it is the reason this guide exists. Every layered TTS tool on GitHub (apps over engines over weights) must be licensed layer by layer before a single commercial minute is rendered, and the layer that produces the sound is the layer that binds the output.
The dependable shelf
For a builder who wants to depend on something: Chatterbox is the strongest repository profile in the set (MIT code and release, an active vendor behind it, zero-shot cloning, a watermark you should treat as a feature to disclose rather than a bug to strip). Kokoro is the dependable narration engine: Apache-2.0 in both layers, tiny, CPU-friendly, and boring in the way infrastructure should be. Piper's living repo (piper1-gpl) is the offline and embedded standard. Qwen3-TTS and the newer model drops are where the quality frontier moves; their repos are worth watching and their licenses worth reading on the day you adopt, not the day you bookmark. XTTS remains excellent and frozen: adopt it knowing no one is coming to fix it, under a model license that forbids commercial use regardless of the toolkit's MPL-2.0. The checklist below turns the five checks into a repeatable audit you can run on any new repo this query surfaces next month.
Interactive tool
Repository due diligence
Check off each step as you finish it. Your progress stays on this page.
0 of 6 complete
0% done
Related guides
The shortlist does not end here. These related guides cover the sibling tools, matchups, and pricing questions closest to this one.
What is the best ElevenLabs alternative on GitHub?
For cloning, Chatterbox (MIT, actively maintained by Resemble AI). For narration, Kokoro (Apache-2.0, runs on CPU). For offline and embedded speech, Piper's current home at piper1-gpl. Best means the repository profile (license in both layers, release cadence, maintenance), not the demo quality alone.
Why do you ignore star counts?
Stars measure cumulative attention, including attention from years ago. They do not move when a repo is archived (Piper's original), when a weights license forbids your use (F5-TTS), or when a company folds (Coqui). Release dates, license files, and issue activity measure the things that determine whether your project survives contact with the repo.
Is VoiceStudio free for commercial use?
The app is AGPL-3.0, but its default engine, OmniVoice, is CC-BY-NC non-commercial, and the engine's license governs the audio you produce. Commercial use requires a differently licensed engine or a license arrangement. Check every layer of any TTS app before rendering commercial work.
What happened to Piper on GitHub?
The original repository was archived on October 6, 2025. Development continues as piper1-gpl under the Open Home Foundation, which is also where Home Assistant's speech ecosystem points. Bookmarks and tutorials citing the old repo are pointing at a frozen snapshot.
Can I use XTTS v2 in a commercial product?
No. XTTS v2 is released under the Coqui Public Model License, which is non-commercial, whatever the surrounding toolkit's license says. With Coqui shut down, there is no one to sell you an exception. For commercial cloning from GitHub, Chatterbox's MIT release is the clean path.
Photo credits: Godfrey Atima (Pexels). Photos are served from this site; follow the credit to the library page for the original.