engineering/uncloseai.com
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

text-to-speech.html: F5-TTS is the live engine, not Qwen3-TTS

Browse source 
Updates marketing copy to reflect actual production state: F5-TTS replaced
Qwen3-TTS as the default engine after head-to-head benchmarking showed
faster inference and cleaner clones on the same reference clips. Engine
count goes from 5 to 6 (F5-TTS additive). Qwen3-TTS remains wired up but
marked Self-host. Updates Makefile snippets to F5 targets, and the sidebar
nav now links F5-TTS at the top of the engine list.
This commit is contained in:
russell@unturf.com 2026-05-25 06:57:40 -04:00
parent bbfeff4b60
commit 535f4ce027
No known key found for this signature in database
1 changed files with 19 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

34
public/text-to-speech.html
View file

@ -25,7 +25,7 @@
<meta name="theme-color" content="#43a047">
<meta name="color-scheme" content="light dark">
<title>Open-Source Text-to-Speech | uncloseai-speech | uncloseai.com</title>
<meta name="description" content="Self-hostable OpenAI-compatible text-to-speech API with 5 TTS engines, 42+ built-in cloned voices, and zero API keys. Default: Qwen3-TTS with voice cloning.">
<meta name="description" content="Self-hostable OpenAI-compatible text-to-speech API with 6 TTS engines, 42 built-in cloned voices, and zero API keys. Default: F5-TTS flow-matching voice cloning.">
<link rel="stylesheet" href="/css/pico.classless.min.css">
<link rel="stylesheet" href="/css/chunkfive/stylesheet.css" type="text/css" charset="utf-8" />
@ -176,6 +176,7 @@
<li><a href="/browser-toys.html">Browser Toys</a></li>
<li><a href="/inference.html">Inference Setup</a></li>
<li><a href="/text-to-speech.html" class="active">Text-to-Speech</a></li>
<li><a href="/text-to-speech.html#f5-tts" style="padding-left:2em">F5-TTS &mdash; Live</a></li>
<li><a href="/tts/voice-cloning.html" style="padding-left:2em">Qwen3-TTS</a></li>
<li><a href="/tts/fast-synthesis.html" style="padding-left:2em">Piper TTS</a></li>
<li><a href="/tts/hd-cloning.html" style="padding-left:2em">XTTS v2</a></li>
@ -216,15 +217,15 @@
<ul>
<li><strong>Zero API Keys:</strong> No registration, no tracking, no rate limits on your own infrastructure</li>
<li><strong>OpenAI-Compatible:</strong> Drop-in replacement &mdash; change one URL and you're running</li>
<li><strong>Five TTS Engines:</strong> Each with different strengths, all behind the same API</li>
<li><strong>42+ Built-In Cloned Voices:</strong> With 10-language support</li>
<li><strong>Six TTS Engines:</strong> Each with different strengths, all behind the same API</li>
<li><strong>42 Built-In Cloned Voices:</strong> 41 from LibriSpeech public domain + 1 self-recorded, 10-language support</li>
<li><strong>Self-Hostable:</strong> Docker compose, Makefile-driven, runs on your hardware</li>
<li><strong>AGPL v3 Licensed:</strong> Keeps TTS libre forever</li>
</ul>
<h2>What's Live Right Now</h2>
<p>Our public endpoint runs the <a href="/tts/voice-cloning.html"><strong>Voice Cloning engine</strong></a> &mdash; 42 distinct human voices, 41 cloned from the LibriSpeech public domain corpus plus the operator's own self-recorded voice (<code>foxhop</code>, voice 42). Speaks 10 languages natively, first audio arrives in under 100 milliseconds.</p>
<p>Our public endpoint runs <a href="#f5-tts"><strong>F5-TTS</strong></a> &mdash; a flow-matching voice-cloning engine that benchmarks faster and produces cleaner clones than Qwen3-TTS on the same reference audio. 42 distinct human voices: 41 cloned from the LibriSpeech public domain corpus plus the operator's own self-recorded voice (<code>foxhop</code>, voice 42). Speaks 10 languages natively, first audio arrives in under 100 milliseconds.</p>
<div class="tts-demo">
<pre><code class="bash"># Female voice
@ -280,7 +281,7 @@ client.audio.speech.create(
client.audio.speech.create(
voice="atlas",
input="Five engines, one API. Self-host it, clone any voice, speak ten languages. No vendor lock-in, no API keys, no limits."
input="Six engines, one API. Self-host it, clone any voice, speak ten languages. No vendor lock-in, no API keys, no limits."
).stream_to_file("atlas.mp3")</code></pre>
<h3>Browse Voices &amp; Models</h3>
@ -289,12 +290,15 @@ client.audio.speech.create(
<p><a href="https://speech.ai.unturf.com/v1/voices" target="_blank"><strong>&rarr; See all voices</strong></a> &nbsp; <a href="https://speech.ai.unturf.com/v1/models" target="_blank"><strong>&rarr; See all engines</strong></a></p>
<h2>The Five Engines</h2>
<h2>The Six Engines</h2>
<p>Five TTS engines, each rescued from a different corner of open source, all running behind the same OpenAI-compatible API. Our public endpoint runs Qwen3-TTS. The other four are ready for anyone who clones the repo.</p>
<p>Six TTS engines, each rescued from a different corner of open source, all running behind the same OpenAI-compatible API. Our public endpoint runs F5-TTS. The other five are ready for anyone who clones the repo.</p>
<h3><a href="/tts/voice-cloning.html">Qwen3-TTS</a> &mdash; <mark>Live</mark></h3>
<p>42+ cloned voices, 10 languages, voice cloning from 3-second samples. 1.7 billion parameters, 97ms first-packet latency. This is what's running on our public endpoint right now.</p>
<h3 id="f5-tts">F5-TTS &mdash; <mark>Live</mark></h3>
<p>Flow-matching voice cloning, 42 voices, 10 languages. Model is F5-TTS_v1 + Vocos vocoder (~1.5GB), MIT-licensed, sourced from <code>SWivid/F5-TTS</code> on Hugging Face. Empirically faster inference and cleaner clones than Qwen3-TTS on the same reference clips. This is what's running on our public endpoint right now &mdash; called as <code>"model":"tts-1-f5"</code>.</p>
<h3><a href="/tts/voice-cloning.html">Qwen3-TTS</a> &mdash; <small>Self-host</small></h3>
<p>42 cloned voices, 10 languages, voice cloning from 3-second samples. 1.7 billion parameters, 97ms first-packet latency. Still wired up in the repo (<code>"model":"tts-1-qwen"</code>); F5-TTS replaced it as our default after head-to-head benchmarking.</p>
<h3><a href="/tts/fast-synthesis.html">Piper TTS</a> &mdash; <small>Self-host</small></h3>
<p>100+ English voices, CPU-only, ONNX runtime. The fastest engine in the dumpster &mdash; built for high-volume batch jobs and real-time applications where latency matters most.</p>
@ -310,22 +314,22 @@ client.audio.speech.create(
<h2>Self-Hosting</h2>
<p>Clone the repo, deploy, and you have your own production TTS API. All five engines are included &mdash; enable whichever ones you need.</p>
<p>Clone the repo, deploy, and you have your own production TTS API. All six engines are included &mdash; enable whichever ones you need.</p>
<pre><code class="bash">git clone https://git.unturf.com/engineering/unturf/uncloseai-speech.git
cd uncloseai-speech
# Deploy with GPU (Qwen3-TTS, default)
# Deploy with GPU (F5-TTS, default)
make deploy
# Or CPU-only (works anywhere, slower)
make deploy-cpu
# Download 42+ cloned voice samples from LibriSpeech
make voices-qwen
# Download 42 cloned voice samples (LibriSpeech + foxhop self-recorded)
make voices-f5
# Test it (Qwen3-TTS model ~3.4GB downloads automatically on first use)
make test
# Test it (F5-TTS model ~1.5GB downloads automatically on first use)
make test-f5
# Enable additional engines
make voices-piper # Piper TTS