Simplify text-to-speech page content
- Add /v1/models endpoint example with jq pretty printing - Fix /v1/voices endpoint to use jq . for pretty printing - Remove 'Learn ML By Doing' section, link to repo README instead - Remove 'For Machine Learning Learners' section - Remove 'Contribute Voices' section - Remove 'Why AGPL v3' section - Remove 'What's Next' roadmap section - Remove MinIO infrastructure goals reference
This commit is contained in:
parent
91859d1807
commit
12b10b01cd
1 changed files with 9 additions and 87 deletions
|
|
@ -155,24 +155,19 @@ client.audio.speech.create(
|
|||
|
||||
<p>Extended <code>/v1/voices</code> endpoint shows all 227 voices with engine metadata:</p>
|
||||
|
||||
<pre><code class="bash">curl https://speech.ai.unturf.com/v1/voices | jq</code></pre>
|
||||
<pre><code class="bash">curl https://speech.ai.unturf.com/v1/voices | jq .</code></pre>
|
||||
|
||||
<h2 id="self-hosting">Self-Hosting: Learn ML By Doing</h2>
|
||||
<h3 id="discover-models">Discover Models</h3>
|
||||
|
||||
<p>This isn't just a TTS server - it's a <strong>learning platform</strong> for understanding production ML infrastructure:</p>
|
||||
<p>Standard OpenAI <code>/v1/models</code> endpoint shows available TTS models:</p>
|
||||
|
||||
<h3 id="what-youll-learn">What You'll Learn</h3>
|
||||
<pre><code class="bash">curl https://speech.ai.unturf.com/v1/models | jq .</code></pre>
|
||||
|
||||
<ul>
|
||||
<li><strong>Model Loading & Inference:</strong> How ONNX, PyTorch, and custom pipelines work in production</li>
|
||||
<li><strong>Voice Routing Logic:</strong> Auto-detection, caching, and O(1) lookups with 227+ voices</li>
|
||||
<li><strong>Audio Processing:</strong> FFmpeg pipelines, format conversion, sample rate handling</li>
|
||||
<li><strong>API Design:</strong> OpenAI compatibility, FastAPI, streaming responses</li>
|
||||
<li><strong>Docker & Deployment:</strong> Multi-stage builds, GPU vs CPU images, remote deployment via Makefile</li>
|
||||
<li><strong>Model Integration:</strong> See how we integrated 4 different TTS engines with completely different APIs</li>
|
||||
</ul>
|
||||
<h2 id="self-hosting">Self-Hosting</h2>
|
||||
|
||||
<h3 id="quick-start">Quick Start (5 Minutes)</h3>
|
||||
<p>For complete installation instructions and documentation, see the <a href="https://git.unturf.com/engineering/unturf/uncloseai-speech" target="_blank">uncloseai-speech repository</a>.</p>
|
||||
|
||||
<h3 id="quick-start">Quick Start</h3>
|
||||
|
||||
<pre><code class="bash"># Clone the repository
|
||||
git clone https://git.unturf.com/engineering/unturf/uncloseai-speech.git
|
||||
|
|
@ -246,32 +241,7 @@ make test</code></pre>
|
|||
<li><strong>Tech:</strong> Apache 2.0 license, 24kHz output</li>
|
||||
</ul>
|
||||
|
||||
<h2 id="for-ml-learners">For Machine Learning Learners</h2>
|
||||
|
||||
<p>The codebase is documented with ML learners in mind:</p>
|
||||
|
||||
<h3 id="code-walkthrough">Code Walkthrough</h3>
|
||||
|
||||
<ul>
|
||||
<li><code>speech.py</code> - Main API server, see how FastAPI + ML models work together</li>
|
||||
<li><code>voice_to_speaker.default.yaml</code> - Voice routing config, understand model-specific parameters</li>
|
||||
<li><code>Makefile</code> - Production deployment patterns, remote Docker orchestration</li>
|
||||
<li><code>docs/MODELS.md</code> - Deep dive into each TTS engine, integration patterns</li>
|
||||
<li><code>docs/CLAUDE.md</code> - Contribution guidelines, architecture decisions</li>
|
||||
</ul>
|
||||
|
||||
<h3 id="learning-path">Suggested Learning Path</h3>
|
||||
|
||||
<ol>
|
||||
<li><strong>Week 1:</strong> Deploy locally, test all 4 engines, read <code>speech.py</code></li>
|
||||
<li><strong>Week 2:</strong> Add a custom Piper voice, understand YAML config structure</li>
|
||||
<li><strong>Week 3:</strong> Clone your own voice with XTTS, study voice cloning code</li>
|
||||
<li><strong>Week 4:</strong> Integrate a new TTS engine from <code>docs/MODELS.md</code> roadmap</li>
|
||||
<li><strong>Week 5:</strong> Deploy to production, measure latency, optimize caching</li>
|
||||
<li><strong>Week 6:</strong> Contribute back - fix a bug, add documentation, improve tests</li>
|
||||
</ol>
|
||||
|
||||
<h2 id="get-involved">Get Involved (Hosting for the Community)</h2>
|
||||
<h2 id="get-involved">Get Involved</h2>
|
||||
|
||||
<p>We run a free public endpoint at <code>https://speech.ai.unturf.com/v1</code> but we need help scaling:</p>
|
||||
|
||||
|
|
@ -292,54 +262,6 @@ make test</code></pre>
|
|||
<li><strong>Build Tools:</strong> Voice editor UI, quality benchmarking, automated testing</li>
|
||||
</ul>
|
||||
|
||||
<h3 id="contribute-voices">Contribute Voices</h3>
|
||||
|
||||
<p>We're building a community voice library:</p>
|
||||
|
||||
<ul>
|
||||
<li><strong>Record Samples:</strong> 6-30 seconds, clean audio, diverse accents</li>
|
||||
<li><strong>Public Domain Only:</strong> Your voice, your consent, Apache 2.0 license</li>
|
||||
<li><strong>Attribution:</strong> Get credited in the voice library metadata</li>
|
||||
</ul>
|
||||
|
||||
<h2 id="why-agpl">Why AGPL v3?</h2>
|
||||
|
||||
<p>We use <strong>AGPL v3</strong> (not MIT, not Apache) because we believe in network freedom:</p>
|
||||
|
||||
<blockquote>
|
||||
<p><strong>"If you run this as a network service (API), you must provide the source code to your users."</strong></p>
|
||||
</blockquote>
|
||||
|
||||
<p>This ensures:</p>
|
||||
<ul>
|
||||
<li>No proprietary forks can lock users into closed platforms</li>
|
||||
<li>Improvements benefit the entire community</li>
|
||||
<li>TTS remains free as in freedom, not just free as in beer</li>
|
||||
<li>Even SaaS providers must contribute back</li>
|
||||
</ul>
|
||||
|
||||
<p><strong>For self-hosting:</strong> Do whatever you want. The AGPL only kicks in if you provide it as a network service to others.</p>
|
||||
|
||||
<h2 id="roadmap">What's Next?</h2>
|
||||
|
||||
<h3 id="integration-pipeline">Integration Pipeline</h3>
|
||||
|
||||
<ol>
|
||||
<li><strong>StyleTTS2</strong> ⭐⭐⭐⭐⭐ - State-of-the-art prosody and naturalness</li>
|
||||
<li><strong>Fish Speech</strong> ⭐⭐⭐⭐ - Fast, modern, excellent multilingual support</li>
|
||||
<li><strong>Chatterbox</strong> ⭐⭐⭐⭐ - Emotion control, 23 languages, zero-shot cloning</li>
|
||||
</ol>
|
||||
|
||||
<h3 id="infrastructure-goals">Infrastructure Goals</h3>
|
||||
|
||||
<ul>
|
||||
<li>Binary mirror network (MinIO) for voice model distribution</li>
|
||||
<li>Automated quality testing framework</li>
|
||||
<li>Voice library with 100+ diverse accents</li>
|
||||
<li>Real-time streaming synthesis (WebSockets)</li>
|
||||
<li>Multi-region deployment with automatic failover</li>
|
||||
</ul>
|
||||
|
||||
<h2 id="resources">Resources</h2>
|
||||
|
||||
<ul>
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue