From 535f4ce027aef45d0636d9f8d5ffc4071f5b7d22 Mon Sep 17 00:00:00 2001 From: "russell@unturf.com" Date: Mon, 25 May 2026 06:57:40 -0400 Subject: [PATCH] text-to-speech.html: F5-TTS is the live engine, not Qwen3-TTS 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. --- public/text-to-speech.html | 34 +++++++++++++++++++--------------- 1 file changed, 19 insertions(+), 15 deletions(-) diff --git a/public/text-to-speech.html b/public/text-to-speech.html index a4d1725..24e36be 100644 --- a/public/text-to-speech.html +++ b/public/text-to-speech.html @@ -25,7 +25,7 @@ Open-Source Text-to-Speech | uncloseai-speech | uncloseai.com - + @@ -176,6 +176,7 @@
  • Browser Toys
  • Inference Setup
  • Text-to-Speech
    • +
  • F5-TTS — Live
  • Qwen3-TTS
  • Piper TTS
  • XTTS v2
    • @@ -216,15 +217,15 @@

    What's Live Right Now

    -

    Our public endpoint runs the Voice Cloning engine — 42 distinct human voices, 41 cloned from the LibriSpeech public domain corpus plus the operator's own self-recorded voice (foxhop, voice 42). Speaks 10 languages natively, first audio arrives in under 100 milliseconds.

    +

    Our public endpoint runs F5-TTS — 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 (foxhop, voice 42). Speaks 10 languages natively, first audio arrives in under 100 milliseconds.

    # 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")

    Browse Voices & Models

    @@ -289,12 +290,15 @@ client.audio.speech.create(

    → See all voices   → See all engines

    -

    The Five Engines

    +

    The Six Engines

    -

    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.

    +

    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.

    -

    Qwen3-TTS — Live

    -

    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.

    +

    F5-TTS — Live

    +

    Flow-matching voice cloning, 42 voices, 10 languages. Model is F5-TTS_v1 + Vocos vocoder (~1.5GB), MIT-licensed, sourced from SWivid/F5-TTS 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 — called as "model":"tts-1-f5".

    + +

    Qwen3-TTS — Self-host

    +

    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 ("model":"tts-1-qwen"); F5-TTS replaced it as our default after head-to-head benchmarking.

    Piper TTS — Self-host

    100+ English voices, CPU-only, ONNX runtime. The fastest engine in the dumpster — built for high-volume batch jobs and real-time applications where latency matters most.

    @@ -310,22 +314,22 @@ client.audio.speech.create(

    Self-Hosting

    -

    Clone the repo, deploy, and you have your own production TTS API. All five engines are included — enable whichever ones you need.

    +

    Clone the repo, deploy, and you have your own production TTS API. All six engines are included — enable whichever ones you need.

    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