<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Cengizhan Kuruderioğlu</title>
    <description>The latest articles on DEV Community by Cengizhan Kuruderioğlu (@cengizhan_kuruderiolu_bb).</description>
    <link>https://dev.to/cengizhan_kuruderiolu_bb</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4257092%2F861d5064-bdd4-4965-99ab-a1d04633f030.png</url>
      <title>DEV Community: Cengizhan Kuruderioğlu</title>
      <link>https://dev.to/cengizhan_kuruderiolu_bb</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/cengizhan_kuruderiolu_bb"/>
    <language>en</language>
    <item>
      <title>Detect emotion in speech with a REST API in 5 minutes</title>
      <dc:creator>Cengizhan Kuruderioğlu</dc:creator>
      <pubDate>Sat, 21 Nov 2026 10:05:52 +0000</pubDate>
      <link>https://dev.to/cengizhan_kuruderiolu_bb/detect-emotion-in-speech-with-a-rest-api-in-5-minutes-5gm4</link>
      <guid>https://dev.to/cengizhan_kuruderiolu_bb/detect-emotion-in-speech-with-a-rest-api-in-5-minutes-5gm4</guid>
      <description>&lt;p&gt;Words and tone often disagree. "Great, thanks" can be warm or sarcastic. A speech-to-text pipeline gives you the words. In this short tutorial you will add the tone using a REST API that estimates emotion from a voice clip.&lt;/p&gt;

&lt;p&gt;I'm the author of the API (Ada, by Adanory), so keep that in mind. I'll stick to what it does and where it falls short.&lt;/p&gt;

&lt;h2&gt;
  
  
  What you get
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Four labels: &lt;code&gt;neutral&lt;/code&gt;, &lt;code&gt;angry&lt;/code&gt;, &lt;code&gt;happy&lt;/code&gt;, &lt;code&gt;sad&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;A confidence and a score per label&lt;/li&gt;
&lt;li&gt;No transcript, so it runs next to whatever speech-to-text you use&lt;/li&gt;
&lt;li&gt;Free tier: 30 analyses once, no card. One request is one analysis.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  1. Get an API key
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Create an account at &lt;a href="https://chat.adanory.xyz/login" rel="noopener noreferrer"&gt;https://chat.adanory.xyz/login&lt;/a&gt; and verify your e-mail.&lt;/li&gt;
&lt;li&gt;Open Settings, choose the API tab and create a key. It's shown once.&lt;/li&gt;
&lt;li&gt;Put it in an environment variable:
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="kd"&gt;export&lt;/span&gt; &lt;span class="nv"&gt;ADA_API_KEY&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s1"&gt;'your-key-here'&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Never put the key in browser code or a public repo.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Prepare a WAV clip
&lt;/h2&gt;

&lt;p&gt;The API accepts WAV files that are 0.5 to 15 seconds, mono, 16 kHz, 16-bit PCM. Convert any recording with ffmpeg and trim to 5 seconds:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;ffmpeg -i input.m4a -ss 00:00:02 -t 5 -ac 1 -ar 16000 -c:a pcm_s16le sample.wav
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Or download the synthetic example: &lt;a href="https://adanory.xyz/static/voice-analysis-sample.wav" rel="noopener noreferrer"&gt;https://adanory.xyz/static/voice-analysis-sample.wav&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  3. Call it with curl
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;curl --fail-with-body --max-time 90 \
  &lt;span class="s1"&gt;'https://adanory.xyz/api/v1/emotion?model=light'&lt;/span&gt; \
  -H &lt;span class="s2"&gt;"Authorization: Bearer ${ADA_API_KEY}"&lt;/span&gt; \
  -H &lt;span class="s1"&gt;'Content-Type: audio/wav'&lt;/span&gt; \
  --data-binary &lt;span class="s1"&gt;'@sample.wav'&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The response looks like this (illustrative values):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight json"&gt;&lt;code&gt;&lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="nl"&gt;"emotion"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="nl"&gt;"label"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"happy"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="nl"&gt;"confidence"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;0.82&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="nl"&gt;"confident"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="nl"&gt;"probabilities"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"neutral"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;0.08&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"angry"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;0.04&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"happy"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;0.82&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"sad"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;0.06&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"stored_for_training"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;false&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"analyses_left"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;29&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"model"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"light"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"model_name"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"Ada Light"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="nl"&gt;"duration_sec"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mf"&gt;2.5&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;analyses_left&lt;/code&gt; is your remaining balance and &lt;code&gt;stored_for_training&lt;/code&gt; is &lt;code&gt;false&lt;/code&gt; unless you opted in to training in Settings.&lt;/p&gt;

&lt;h2&gt;
  
  
  4. Call it from Python
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="n"&gt;base64&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;os&lt;/span&gt;
&lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="n"&gt;pathlib&lt;/span&gt; &lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="nc"&gt;Path&lt;/span&gt;
&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="n"&gt;requests&lt;/span&gt;

&lt;span class="n"&gt;response&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;requests&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;post&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;
    &lt;span class="s2"&gt;"https://adanory.xyz/api/v1/emotion"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="n"&gt;headers&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="s2"&gt;"Authorization"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;f&lt;/span&gt;&lt;span class="s2"&gt;"Bearer {os.environ['ADA_API_KEY']}"&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="n"&gt;json&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;&lt;span class="s2"&gt;"audio_b64"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;base64&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;b64encode&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nc"&gt;Path&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"sample.wav"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;read_bytes&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;decode&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"ascii"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
          &lt;span class="s2"&gt;"model"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"light"&lt;/span&gt;&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="n"&gt;timeout&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="mi"&gt;90&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
&lt;span class="p"&gt;)&lt;/span&gt;
&lt;span class="n"&gt;response&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;raise_for_status&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;span class="n"&gt;result&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;response&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;json&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;span class="k"&gt;print&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;result&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"emotion"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"label"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;result&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"emotion"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="s2"&gt;"confidence"&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Errors come back as an HTTP status plus a JSON &lt;code&gt;error&lt;/code&gt; string, for example 402 when the balance is empty and 400 when the WAV is not 16 kHz mono 16-bit. Send an &lt;code&gt;Idempotency-Key&lt;/code&gt; header when you retry so a repeated successful request isn't charged twice.&lt;/p&gt;

&lt;h2&gt;
  
  
  5. Choose a model
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;pro&lt;/code&gt; (default): English-focused. On public benchmarks with unseen speakers it reaches about 92% on SAVEE and 83% on TESS.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;light&lt;/code&gt;: faster.&lt;/li&gt;
&lt;li&gt;
&lt;code&gt;exp&lt;/code&gt;: experimental Turkish. About 58% on TurEV-DB, where chance is 25%.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Add &lt;code&gt;X-Adanory-Strict-Model: true&lt;/code&gt; if you want an error instead of a silent fallback when a model is unavailable.&lt;/p&gt;

&lt;h2&gt;
  
  
  Honest limitations
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Four labels are a coarse view of emotion.&lt;/li&gt;
&lt;li&gt;Benchmark numbers come from acted speech. Phone audio, noise and accents will differ, so test with your own clips.&lt;/li&gt;
&lt;li&gt;Clips are limited to 15 seconds; split long calls into utterance-sized windows.&lt;/li&gt;
&lt;li&gt;It's an estimate that can be wrong. It is not a medical or diagnostic tool and not for decisions about people.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Next steps
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;API docs: &lt;a href="https://adanory.xyz/docs?utm_source=devto&amp;amp;utm_medium=organic&amp;amp;utm_campaign=launch" rel="noopener noreferrer"&gt;https://adanory.xyz/docs?utm_source=devto&amp;amp;utm_medium=organic&amp;amp;utm_campaign=launch&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Pricing: &lt;a href="https://adanory.xyz/pricing" rel="noopener noreferrer"&gt;https://adanory.xyz/pricing&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Guide for voice agents: &lt;a href="https://adanory.xyz/emotion-detection-for-voice-ai-agents" rel="noopener noreferrer"&gt;https://adanory.xyz/emotion-detection-for-voice-ai-agents&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Try the chat free, no account needed for the first 3 messages: &lt;a href="https://chat.adanory.xyz/?utm_source=devto&amp;amp;utm_medium=organic&amp;amp;utm_campaign=launch" rel="noopener noreferrer"&gt;https://chat.adanory.xyz/?utm_source=devto&amp;amp;utm_medium=organic&amp;amp;utm_campaign=launch&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;If you try it, I'd like to hear where it misreads you.&lt;/p&gt;

</description>
      <category>api</category>
      <category>python</category>
      <category>ai</category>
      <category>tutorial</category>
    </item>
  </channel>
</rss>
