Create app.py

app.py

@@ -0,0 +1,82 @@
+import gradio as gr
+import tempfile
+import soundfile as sf
+
+from pocket_tts import TTSModel
+
+# -------------------------------------------------
+# Load Pocket-TTS model ONCE (HF CPU-safe)
+# -------------------------------------------------
+model = TTSModel.load_model()
+
+# -------------------------------------------------
+# All official Kyutai voices (from kyutai/tts-voices)
+# -------------------------------------------------
+VOICES = {
+    "Alba – Casual": "hf://kyutai/tts-voices/alba-mackenna/casual.wav",
+    "Alba – Announcer": "hf://kyutai/tts-voices/alba-mackenna/announcer.wav",
+    "Alba – Merchant": "hf://kyutai/tts-voices/alba-mackenna/merchant.wav",
+    "Alba – A Moment By": "hf://kyutai/tts-voices/alba-mackenna/a-moment-by.wav",
+}
+
+# -------------------------------------------------
+# TTS generation (voice state per request)
+# -------------------------------------------------
+def generate_tts(text, voice_name):
+    if not text.strip():
+        return None
+
+    voice_path = VOICES[voice_name]
+
+    # IMPORTANT:
+    # Create voice state per request
+    voice_state = model.get_state_for_audio_prompt(voice_path)
+
+    audio = model.generate_audio(
+        voice_state=voice_state,
+        text=text,
+    )
+
+    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
+    sf.write(tmp.name, audio, samplerate=24000)
+
+    return tmp.name
+
+
+# -------------------------------------------------
+# Gradio UI
+# -------------------------------------------------
+with gr.Blocks(title="Pocket TTS – Kyutai Voices") as demo:
+    gr.Markdown(
+        """
+        # 🗣️ Pocket TTS (Kyutai)
+        CPU-only • Hugging Face native • All official voices  
+        """
+    )
+
+    voice_select = gr.Dropdown(
+        choices=list(VOICES.keys()),
+        value="Alba – Casual",
+        label="Voice"
+    )
+
+    text_input = gr.Textbox(
+        label="Text",
+        lines=4,
+        placeholder="Type something and generate speech..."
+    )
+
+    generate_btn = gr.Button("Generate")
+
+    audio_output = gr.Audio(
+        label="Output",
+        autoplay=False
+    )
+
+    generate_btn.click(
+        fn=generate_tts,
+        inputs=[text_input, voice_select],
+        outputs=audio_output
+    )
+
+demo.launch()