Skip to main content
PromptQuorum
Home/Power Local LLM/openai-edge-tts Review: A Free OpenAI-Compatible TTS Server
Voice, Speech & Multimodal

openai-edge-tts Review: A Free OpenAI-Compatible TTS Server

Β·10 min readΒ·By Hans Kuepper Β· Founder of PromptQuorum, multi-model AI dispatch tool Β· PromptQuorum

openai-edge-tts is a free, open-source, self-hosted API server that exposes an /v1/audio/speech endpoint matching OpenAI's text-to-speech API format, but it does not synthesize speech locally. Every request is forwarded through the edge-tts library to Microsoft Edge's free online "Read Aloud" voices over the internet, so text you send to it leaves your machine and reaches Microsoft's service, even though the server process itself runs on hardware you control.

openai-edge-tts (github.com/travisvn/openai-edge-tts) is a free, open-source, self-hosted API server, created by developer travisvn, that exposes an /v1/audio/speech endpoint matching OpenAI's text-to-speech API format. It does not run a local speech model β€” it forwards each request through the edge-tts Python library to Microsoft Edge's free online "Read Aloud" voices, then returns the audio in an OpenAI-compatible response. This review covers exactly what it proxies to, how to install it, and who it fits.

Key Takeaways

  • openai-edge-tts (github.com/travisvn/openai-edge-tts) is a free, open-source, self-hosted API server β€” not a local speech model
  • Created by developer travisvn; repository created October 9, 2024
  • GPL-3.0 licensed, confirmed via the GitHub repository's license field
  • Exposes /v1/audio/speech, matching OpenAI's text-to-speech API request/response shape, so existing OpenAI-TTS client code can point at it with a changed base URL
  • Synthesis is proxied to Microsoft Edge's free online "Read Aloud" voices via the edge-tts Python library β€” no local speech model runs on your hardware
  • Over 2,100 GitHub stars and 316 forks as of this review

πŸ“ In One Sentence

openai-edge-tts is a free, open-source (GPL-3.0) API server created by travisvn that exposes an OpenAI-compatible /v1/audio/speech endpoint, but instead of running a local speech model, it proxies every request to Microsoft Edge's free online voices via the edge-tts library.

πŸ’¬ In Plain Terms

It is a small server you run yourself (usually in Docker) that makes text-to-speech tools built for OpenAI's API work against a free voice source instead β€” but the actual "speaking" happens on Microsoft's servers, not on your machine. If privacy from Microsoft matters for your text, this tool does not provide it; if the cost of the OpenAI/Azure/ElevenLabs TTS API is the problem you are solving, it does.

πŸ“ŒNote: This review is based on openai-edge-tts's own GitHub repository, README, and release notes. It does not claim PromptQuorum has run hands-on latency or quality benchmarks against the underlying Microsoft Edge voices.

What Is openai-edge-tts?

openai-edge-tts is a self-hosted API server that mimics OpenAI's text-to-speech API, so tools built to call OpenAI, Azure AI Speech, or ElevenLabs can be pointed at it instead β€” without a per-request bill. It does this by wrapping the edge-tts library, which itself talks to Microsoft Edge's free, browser-based "Read Aloud" voice service over the internet.

  • Product type: a self-hosted API server (Docker or Python), not a downloadable end-user app and not a local speech model
  • Creator: travisvn, a solo developer; the GitHub repository is hosted at github.com/travisvn/openai-edge-tts
  • Repository created October 9, 2024, per GitHub's repository metadata
  • License: GPL-3.0, confirmed via the GitHub repository's license metadata; the README also states enterprise/commercial deployments should contact the author directly
  • Locality: hybrid, not local β€” the server process is self-hosted, but the actual text-to-speech synthesis is forwarded to Microsoft Edge's online voice service, not computed on your own hardware
  • Scale: over 2,100 GitHub stars and 316 forks as of this review

Project History and Version Milestones

The GitHub repository was created on October 9, 2024, and its most significant public release milestone is v2.0.0, published December 28, 2024, which added markdown filtering, beta ElevenLabs/Azure AI Speech endpoint compatibility, and simplified setup.

  1. 1
    October 9, 2024: Repository created
    Why it matters: Marks the start of the project, per GitHub's repository metadata.
  2. 2
    v2.0.0 β€” December 28, 2024: Markdown filtering, expanded API support, simplified setup
    Why it matters: Added optional markdown filtering on output text, beta drop-in compatibility with ElevenLabs and Azure AI Speech endpoints (alongside the existing OpenAI-format endpoint), and made the `/v1` route prefix optional, per the official GitHub release notes.

What Does openai-edge-tts Actually Do?

openai-edge-tts receives a text-to-speech request in OpenAI's API format, maps the requested voice and settings to an equivalent Microsoft Edge voice, fetches the synthesized audio from Microsoft's online service via the edge-tts library, and returns it in the format the client expects.

  • OpenAI-compatible endpoint: /v1/audio/speech (the /v1 prefix is optional as of v2.0.0), matching OpenAI's TTS request/response shape so existing client code needs only a base-URL change
  • Voice mapping: OpenAI voice names (alloy, echo, fable, onyx, nova, shimmer) map to specific edge-tts voices by default; the README also documents selecting any edge-tts voice directly, bypassing the OpenAI-name mapping
  • Beta compatibility layers: drop-in endpoints mimicking ElevenLabs and Azure AI Speech APIs, in addition to the primary OpenAI-format endpoint, introduced in v2.0.0
  • Audio formats: mp3, opus, aac, flac, wav, and pcm output, per the README
  • Streaming: Server-Sent Events (SSE) streaming with base64-encoded audio chunks, for clients that want incremental audio rather than waiting for a full file
  • Speed control: adjustable playback speed from 0.25x to 4.0x
  • Optional markdown filtering: strips markdown syntax from input text before synthesis by default, since source text (for example, from an LLM response) often contains markdown that should not be read aloud literally; this can be disabled via a REMOVE_FILTER environment variable
  • Configuration: an .env file sets the API key, port (default 5050), default voice, default speed, and default language

Usage Examples: Two Ways to Use openai-edge-tts

These are concrete workflows built from the project's documented features above.

Platform, Pricing, and Licensing

Platform

What openai-edge-tts states:
A self-hosted API server (Docker or Python) β€” no GUI, no downloadable end-user app; runs on any host that can run Docker or Python.

Cost

What openai-edge-tts states:
Free and open source for personal use. The README states enterprise/commercial deployments should contact the author (travisvn) directly.

Licensing

What openai-edge-tts states:
GPL-3.0, confirmed via the GitHub repository's license metadata.

Install method

What openai-edge-tts states:
Docker (docker run or docker-compose) is the documented primary path; a Python virtual-environment + pip path is also documented for running without Docker.

Verify current licensing and commercial-use terms directly in the GitHub repository before deploying this at organizational scale, since the README's enterprise-contact note is a stated intent, not a separate published commercial license text this review could independently verify.

openai-edge-tts vs. Piper TTS

openai-edge-tts and Piper TTS solve the same "free OpenAI-compatible TTS endpoint" problem in opposite ways: one proxies to a free cloud voice service, the other runs a real neural TTS model entirely on your own hardware. They get compared because both show up in the same "how do I add free TTS to my self-hosted AI stack" searches.

Aspect
openai-edge-tts
Piper TTS
Where synthesis happensMicrosoft's Edge voice service, over the networkEntirely on your own CPU, no network call
LocalityHybrid β€” self-hosted server, cloud-dependent synthesisFully local β€” works offline after model download
Voice quality/styleMicrosoft Edge's commercial-grade online voices, many languagesSmaller neural voice models, quality varies by voice, per-voice download
Dependency riskDepends on Microsoft continuing to offer this free serviceDepends only on the downloaded model file continuing to work
LicenseGPL-3.0MIT

If sending text to Microsoft is acceptable for your use case and you want a wide range of polished voices with no local compute cost, openai-edge-tts is the simpler path. If synthesis must stay entirely on hardware you control, Piper TTS (or another genuinely local engine) is the correct choice, not this tool.

Who Should Use openai-edge-tts?

openai-edge-tts fits developers who want to eliminate the per-request cost of a commercial TTS API without giving up an OpenAI-shaped integration, and who are comfortable with text being sent to Microsoft's Edge voice service.

What openai-edge-tts Is Not Good For

openai-edge-tts is not a good fit if genuinely local, offline, or privacy-sensitive speech synthesis is a requirement.

  • Not offline β€” every synthesis request requires a working internet connection to reach Microsoft's Edge voice service; it will not function on an air-gapped machine
  • Not private in the way a local model is β€” text sent for synthesis reaches Microsoft's servers, so this tool does not keep synthesized text confined to your own hardware
  • Not backed by an official, documented Microsoft API β€” it depends on the same free "Read Aloud" voice mechanism used by the Edge browser, which Microsoft could change or restrict without notice; the README itself does not document a formal rate limit or terms-of-service guarantee for this dependency
  • Not a voice-cloning tool β€” it offers Microsoft Edge's existing preset voices only, with no support for training or cloning a custom voice from your own audio
  • Not commercially licensed by default β€” the README asks enterprise/commercial deployments to contact the author directly rather than assuming the GPL-3.0 terms alone cover that use case

Common Mistakes When Evaluating openai-edge-tts

Most confusion about openai-edge-tts comes from assuming it behaves like a local model because it is self-hosted, or from missing the beta status of its non-OpenAI endpoint modes.

Competitors and Alternatives

openai-edge-tts is most often compared to genuinely local, offline text-to-speech engines that also expose an OpenAI-compatible endpoint β€” its main differentiator is trading true offline privacy for Microsoft Edge's broader, more polished voice selection at zero local compute cost.

Tool
Best known for
Link
Piper TTSFast, fully local, CPU-only neural TTS engine from the Rhasspy projectPiper TTS review
Coqui TTSOpen-source local TTS toolkit supporting voice cloning and many languagesCoqui TTS review
XTTS-v2Local, zero-shot voice-cloning TTS model from CoquiXTTS-v2 review
BarkLocal, transformer-based TTS model with expressive/non-speech audio generationBark review

This list reflects tools commonly compared to openai-edge-tts in the local/self-hosted TTS space, not an independent PromptQuorum ranking β€” verify each tool's current feature set and hardware requirements before choosing.

Frequently Asked Questions

What is openai-edge-tts?

openai-edge-tts (github.com/travisvn/openai-edge-tts) is a free, open-source (GPL-3.0), self-hosted API server that exposes an /v1/audio/speech endpoint matching OpenAI's text-to-speech API, backed by Microsoft Edge's free online voices rather than a local speech model.

Is openai-edge-tts local or does it use the cloud?

It is hybrid, not fully local. The server itself is self-hosted, but every text-to-speech request is proxied over the internet to Microsoft Edge's free online "Read Aloud" voice service via the edge-tts library.

Is openai-edge-tts free?

Yes, for personal use. It is free and open source under GPL-3.0. The README states that enterprise/commercial deployments should contact the author, travisvn, directly.

Does openai-edge-tts protect my privacy?

Not in the way a local model does. Because synthesis is proxied to Microsoft's Edge voice service, the text you send is not kept confined to your own hardware β€” do not use it for text you would not want a third-party cloud service to process.

How do I install openai-edge-tts?

The documented quick start is Docker: docker run -d -p 5050:5050 -e API_KEY=your_api_key_here -e PORT=5050 travisvn/openai-edge-tts:latest. A docker-compose setup and a direct Python/pip install are also documented in the GitHub README.

What voices does openai-edge-tts support?

It maps OpenAI's six standard voice names (alloy, echo, fable, onyx, nova, shimmer) to equivalent Microsoft Edge voices by default, and also lets you select any edge-tts voice directly, covering many languages.

Can openai-edge-tts replace ElevenLabs or Azure AI Speech, not just OpenAI?

The project added beta drop-in compatibility endpoints for ElevenLabs and Azure AI Speech in v2.0.0 (December 2024), alongside its primary OpenAI-format endpoint. Verify current behavior against your specific client before relying on this in production.

Who created openai-edge-tts?

A solo developer known as travisvn created and maintains openai-edge-tts. The GitHub repository was created October 9, 2024.

Does openai-edge-tts support voice cloning?

No. It only offers Microsoft Edge's existing preset voices; it has no feature for training or cloning a custom voice from your own audio samples.

Has PromptQuorum independently tested openai-edge-tts's claims?

This review is based on the project's own GitHub repository, README, and release notes, rather than hands-on latency or audio-quality benchmarking by PromptQuorum.

Sources

← Back to Power Local LLM