adding 53 more voices to Kokoro-TTS

app.py

@@ -1,8 +1,8 @@
-# Purpose: One Space that offers four tools/tabs (all exposed as MCP tools):
+# Purpose: One Space that offers six tools/tabs (all exposed as MCP tools):
 #   1) Fetch — extract relevant page content (title, metadata, clean text, hyperlinks)
 #   2) DuckDuckGo Search — compact JSONL search output (short keys to minimize tokens)
 #   3) Python Code Executor — run Python code and capture stdout/errors
-#   4) Kokoro TTS — synthesize speech from text using Kokoro-82M
+#   4) Kokoro TTS — synthesize speech from text using Kokoro-82M with 54 voice options
 #   5) Image Generation - HF serverless inference providers (Default: FLUX.1-Krea-dev)
 #   6) Video Generation - HF serverless inference providers (Default: Wan2.2-T2V-A14B)
 
@@ -457,6 +457,50 @@ _KOKORO_STATE = {
 }
 
 
+def get_kokoro_voices():
+    """Get comprehensive list of available Kokoro voice IDs (54 total)."""
+    try:
+        from huggingface_hub import list_repo_files
+        # Get voice files from the Kokoro repository
+        files = list_repo_files('hexgrad/Kokoro-82M')
+        voice_files = [f for f in files if f.endswith('.pt') and f.startswith('voices/')]
+        voices = [f.replace('voices/', '').replace('.pt', '') for f in voice_files]
+        return sorted(voices) if voices else _get_fallback_voices()
+    except Exception:
+        return _get_fallback_voices()
+
+
+def _get_fallback_voices():
+    """Return comprehensive fallback list of known Kokoro voices (54 total)."""
+    return [
+        # American Female (11 voices)
+        "af_alloy", "af_aoede", "af_bella", "af_heart", "af_jessica", 
+        "af_kore", "af_nicole", "af_nova", "af_river", "af_sarah", "af_sky",
+        # American Male (9 voices)
+        "am_adam", "am_echo", "am_eric", "am_fenrir", "am_liam", 
+        "am_michael", "am_onyx", "am_puck", "am_santa",
+        # British Female (4 voices)
+        "bf_alice", "bf_emma", "bf_isabella", "bf_lily",
+        # British Male (4 voices)
+        "bm_daniel", "bm_fable", "bm_george", "bm_lewis",
+        # European Female/Male (3 voices)
+        "ef_dora", "em_alex", "em_santa",
+        # French Female (1 voice)
+        "ff_siwis",
+        # Hindi Female/Male (4 voices)
+        "hf_alpha", "hf_beta", "hm_omega", "hm_psi",
+        # Italian Female/Male (2 voices)
+        "if_sara", "im_nicola",
+        # Japanese Female/Male (5 voices)
+        "jf_alpha", "jf_gongitsune", "jf_nezumi", "jf_tebukuro", "jm_kumo",
+        # Portuguese Female/Male (3 voices)
+        "pf_dora", "pm_alex", "pm_santa",
+        # Chinese Female/Male (8 voices)
+        "zf_xiaobei", "zf_xiaoni", "zf_xiaoxiao", "zf_xiaoyi",
+        "zm_yunjian", "zm_yunxi", "zm_yunxia", "zm_yunyang"
+    ]
+
+
 def _init_kokoro() -> None:
     """Lazy-initialize Kokoro model and pipelines on first use.
 
@@ -497,26 +541,57 @@ def _init_kokoro() -> None:
     )
 
 
+def List_Kokoro_Voices() -> List[str]:
+    """
+    Get a list of all available Kokoro voice identifiers.
+    
+    This MCP tool helps clients discover the 54 available voice options
+    for the Generate_Speech tool.
+    
+    Returns:
+        List[str]: A list of voice identifiers (e.g., ["af_heart", "am_adam", "bf_alice", ...])
+        
+    Voice naming convention:
+        - First 2 letters: Language/Region (af=American Female, am=American Male, bf=British Female, etc.)
+        - Following letters: Voice name (heart, adam, alice, etc.)
+        
+    Available categories:
+        - American Female/Male (20 voices)
+        - British Female/Male (8 voices) 
+        - European Female/Male (3 voices)
+        - French Female (1 voice)
+        - Hindi Female/Male (4 voices)
+        - Italian Female/Male (2 voices)
+        - Japanese Female/Male (5 voices)
+        - Portuguese Female/Male (3 voices)
+        - Chinese Female/Male (8 voices)
+    """
+    return get_kokoro_voices()
+
+
 def Generate_Speech(  # <-- MCP tool #4 (Generate Speech)
     text: Annotated[str, "The text to synthesize (English)."],
     speed: Annotated[float, "Speech speed multiplier in 0.5–2.0; 1.0 = normal speed."] = 1.25,
-    voice: Annotated[str, "Voice identifier. Example: 'af_heart' (US English, female, Heart)."] = "af_heart",
+    voice: Annotated[str, "Voice identifier from 54 available options. Use List_Kokoro_Voices() to see all choices. Examples: 'af_heart' (US female), 'am_adam' (US male), 'bf_alice' (British female), 'jf_alpha' (Japanese female)."] = "af_heart",
 ) -> Tuple[int, np.ndarray]:
     """
-    Synthesize speech from text using the Kokoro-82M model.
+    Synthesize speech from text using the Kokoro-82M model with 54 voice options.
 
     This function returns raw audio suitable for a Gradio Audio component and is
-    also exposed as an MCP tool (per the latest Hugging Face/Gradio MCP docs, a
-    tool is created for each function wired into your app; docstrings and type
-    hints are used to describe the tool).
+    also exposed as an MCP tool. It supports 54 different voices across multiple
+    languages and accents including American, British, European, Hindi, Italian,
+    Japanese, Portuguese, and Chinese speakers.
 
     Default behavior:
         - Speed defaults to 1.25 (slightly brisk cadence) for clearer, snappier delivery.
+        - Voice defaults to "af_heart" (American Female, Heart voice)
 
     Args:
-        text: The text to synthesize (English).
+        text: The text to synthesize. Works best with English but supports multiple languages.
         speed: Speech speed multiplier in 0.5–2.0; 1.0 = normal speed. Default: 1.25 (slightly brisk).
-        voice: Voice identifier. Example: 'af_heart' (US English, female, Heart).
+        voice: Voice identifier from 54 available options. Use List_Kokoro_Voices() to see all choices.
+               Examples: 'af_heart' (US female), 'am_adam' (US male), 'bf_alice' (British female),
+               'jf_alpha' (Japanese female), 'zf_xiaobei' (Chinese female).
 
     Returns:
         A tuple of (sample_rate_hz, audio_waveform) where:
@@ -524,9 +599,10 @@ def Generate_Speech(  # <-- MCP tool #4 (Generate Speech)
         - audio_waveform: numpy.ndarray float32 mono waveform in range [-1, 1]
 
     Notes:
-        - Requires the 'kokoro' package (>=0.9.4). If unavailable, an error is
-          raised with installation guidance.
+        - Requires the 'kokoro' package (>=0.9.4). If unavailable, an error is raised.
         - Runs on CUDA if available; otherwise CPU.
+        - Supports 54 voices across 9 language/accent categories.
+        - Use List_Kokoro_Voices() MCP tool to discover all available voice options.
     """
     if not text or not text.strip():
         raise gr.Error("Please provide non-empty text to synthesize.")
@@ -640,7 +716,7 @@ CSS_STYLES = """
     /* Place bold tools list on line 2, normal auth note on line 3 (below title) */
     .gradio-container h1::before {
         grid-row: 2;
-        content: "Fetch Webpage | Search DuckDuckGo | Code Interpreter | Kokoro TTS | Image Generation | Video Generation";
+        content: "Fetch Webpage | Search DuckDuckGo | Code Interpreter | Kokoro TTS (54 voices) | Image Generation | Video Generation";
         display: block;
         font-size: 1rem;
         font-weight: 700;
@@ -667,21 +743,28 @@ CSS_STYLES = """
 """
 
 # --- Kokoro TTS tab (text to speech) ---
+available_voices = get_kokoro_voices()
 kokoro_interface = gr.Interface(
     fn=Generate_Speech,
     inputs=[
         gr.Textbox(label="Text", placeholder="Type text to synthesize…", lines=4),
-    gr.Slider(minimum=0.5, maximum=2.0, value=1.25, step=0.1, label="Speed"),
-        gr.Textbox(label="Voice", value="af_heart", placeholder="e.g., af_heart"),
+        gr.Slider(minimum=0.5, maximum=2.0, value=1.25, step=0.1, label="Speed"),
+        gr.Dropdown(
+            label="Voice", 
+            choices=available_voices, 
+            value="af_heart", 
+            info="Select from 54 available voices across multiple languages and accents"
+        ),
     ],
     outputs=gr.Audio(label="Audio", type="numpy"),
     title="Kokoro TTS",
     description=(
-        "<div style=\"text-align:center\">Generate English speech with Kokoro-82M. 30 second max output. Runs on CPU or CUDA if available.</div>"
+        "<div style=\"text-align:center\">Generate speech with Kokoro-82M using 54 different voices. Supports multiple languages and accents. Runs on CPU or CUDA if available.</div>"
     ),
     api_description=(
-        "Synthesize speech from text using Kokoro-82M. Returns (sample_rate, waveform) suitable for playback. "
-        "Parameters: text (str), speed (float 0.5–2.0, default 1.25x), voice (str). "
+        "Synthesize speech from text using Kokoro-82M with 54 voice options. Returns (sample_rate, waveform) suitable for playback. "
+        "Parameters: text (str), speed (float 0.5–2.0, default 1.25x), voice (str from 54 available options). "
+        "Available voices include American/British/European/Hindi/Italian/Japanese/Portuguese/Chinese speakers. "
         "Return the generated media to the user in this format `![Alt text](URL)`"
     ),
     allow_flagging="never",