first commit

.gitignore

@@ -0,0 +1,44 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# Environment
+.env
+.venv
+env/
+venv/
+ENV/
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+*~
+
+# OS
+.DS_Store
+Thumbs.db
+
+# Secrets
+api_keys.json
+secrets.json
+*.key

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2024 gabe-init
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,97 @@
+# ComfyUI ElevenLabs TTS Node
+
+A custom node for ComfyUI that integrates ElevenLabs' text-to-speech API, enabling high-quality voice synthesis within your ComfyUI workflows.
+
+## Features
+
+- Pulls your ElevenLabs voices on demand (cached for 1 hour) and maps names → IDs automatically
+- Built-in support for the shipped models: `eleven_multilingual_v2`, `eleven_english_sts_v2`, `eleven_turbo_v2`
+- Voice tuning controls: stability, similarity boost, style, and speaker boost
+- Text input flexibility: type in the node UI or feed text from another node (`input_text` override)
+- Returns a ComfyUI `AUDIO` dict with waveform tensor + sample rate decoded from ElevenLabs MP3 responses
+
+## Installation
+
+1. Clone this repository into your ComfyUI `custom_nodes` folder:
+```bash
+cd ComfyUI/custom_nodes
+git clone https://github.com/yourusername/ComfyUI-ElevenLabs
+```
+
+2. Install dependencies:
+```bash
+pip install -r requirements.txt
+```
+
+3. Restart ComfyUI.
+
+## Setup
+
+You'll need an ElevenLabs API key (`ELEVENLABS_API_KEY`) set in your environment. Example for a shell session:
+```bash
+export ELEVENLABS_API_KEY=your_key
+```
+The node reads this variable automatically; there is no API key input inside the UI.
+
+## Usage
+
+### Basic Text-to-Speech
+
+1. Add the "ElevenLabs TTS Node" to your workflow.
+2. Type text in the node or wire a string input into `input_text` (wired input takes priority).
+3. Pick a voice from the dropdown (list is fetched from your ElevenLabs account and cached).
+4. Choose a model (defaults to multilingual v2).
+5. Adjust stability, similarity boost, style, and speaker boost as desired.
+6. Connect the audio output to preview or save nodes.
+
+### Inputs
+
+- **text**: Text box content to synthesize.
+- **input_text** (optional): Upstream string input; overrides the text box when connected.
+- **voice**: Dropdown of your ElevenLabs voices (deduped names, numbered when repeated).
+- **model**: One of `eleven_multilingual_v2`, `eleven_english_sts_v2`, `eleven_turbo_v2`.
+- **stability**: 0.0–1.0.
+- **similarity_boost**: 0.0–1.0.
+- **style**: 0.0–1.0.
+- **use_speaker_boost**: Boolean toggle.
+
+### Output
+
+- **AUDIO**: `{"waveform": torch.Tensor(shape: 1 x channels x samples), "sample_rate": int}`
+
+## Notes & Troubleshooting
+
+- Voices are fetched from `https://api.elevenlabs.io/v1/voices` and cached for 1 hour. Missing/expired API keys log an error and return a silent placeholder waveform.
+- Failures in TTS requests or MP3 decoding also return a 1-sample silent waveform and log the error.
+- Some voices or models may depend on your ElevenLabs subscription tier.
+
+## Requirements
+
+- ComfyUI
+- Python packages: requests, torch, numpy
+- ElevenLabs API key
+- Internet connection for API calls
+
+## License
+
+MIT License
+
+Copyright (c) 2024 
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

__init__.py

@@ -0,0 +1,3 @@
+from .node import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS
+
+__all__ = ["NODE_CLASS_MAPPINGS", "NODE_DISPLAY_NAME_MAPPINGS"]

node.py

@@ -0,0 +1,226 @@
+import io
+import os
+import time
+
+import numpy as np
+import requests
+import torch
+from pydub import AudioSegment
+import logging
+
+class ElevenLabsNode:
+    voices_cache = None
+    voices_map = {}
+    models_cache = None
+    last_fetch_time = 0
+    cache_duration = 3600  # Cache duration in seconds (1 hour)
+
+    def __init__(self):
+        pass
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        voices = cls.fetch_elevenlabs_voices()
+        models = cls.fetch_elevenlabs_models()
+        return {
+            "required": {
+                "text": ("STRING", {
+                    "multiline": True,
+                    "default": "Hello, how are you?"
+                }),
+                "voice": (voices,),
+                "model": (models,),
+                "stability": ("FLOAT", {
+                    "default": 0.5,
+                    "min": 0.0,
+                    "max": 1.0,
+                    "step": 0.1
+                }),
+                "similarity_boost": ("FLOAT", {
+                    "default": 0.8,
+                    "min": 0.0,
+                    "max": 1.0,
+                    "step": 0.1
+                }),
+                "style": ("FLOAT", {
+                    "default": 0.0,
+                    "min": 0.0,
+                    "max": 1.0,
+                    "step": 0.1
+                }),
+                "use_speaker_boost": ("BOOLEAN", {
+                    "default": True,
+                }),
+            },
+            "optional": {
+                "input_text": ("STRING", {"forceInput": True}),
+                
+            }
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    FUNCTION = "generate_speech"
+    CATEGORY = "ElevenLabs"
+
+    @classmethod
+    def _get_api_key(cls):
+        api_key = os.getenv("ELEVENLABS_API_KEY", "").strip()
+        if not api_key:
+            logging.error("ELEVENLABS_API_KEY is not set. Please configure it in your environment.")
+            return None
+        return api_key
+
+    @classmethod
+    def fetch_elevenlabs_voices(cls):
+        current_time = time.time()
+        if cls.voices_cache is None or (current_time - cls.last_fetch_time > cls.cache_duration):
+            voice_list = []
+            voice_map = {}
+            api_key = cls._get_api_key()
+            if api_key is None:
+                cls.voices_cache = ["missing_api_key"]
+                cls.voices_map = {}
+                return cls.voices_cache
+
+            # Fetch voices from the API
+            url = "https://api.elevenlabs.io/v1/voices"
+            try:
+                response = requests.get(url, headers={"xi-api-key": api_key})
+                response.raise_for_status()
+                voices = response.json()["voices"]
+                name_counts = {}
+                for voice in voices:
+                    base_name = voice["name"]
+                    count = name_counts.get(base_name, 0) + 1
+                    name_counts[base_name] = count
+                    display_name = base_name if count == 1 else f"{base_name} ({count})"
+                    voice_list.append(display_name)
+                    voice_map[display_name] = voice["voice_id"]
+                cls.voices_cache = voice_list
+                cls.voices_map = voice_map
+                cls.last_fetch_time = current_time
+            except requests.exceptions.RequestException as e:
+                logging.error(f"Error fetching voices: {e}")
+                if cls.voices_cache is None:
+                    cls.voices_cache = ["error_fetching_voices"]
+                cls.voices_map = {}
+        return cls.voices_cache
+
+    @classmethod
+    def fetch_elevenlabs_models(cls):
+        current_time = time.time()
+        if cls.models_cache is None or (current_time - cls.last_fetch_time > cls.cache_duration):
+            cls.models_cache = ["eleven_multilingual_v2", "eleven_english_sts_v2", "eleven_turbo_v2"]
+            cls.last_fetch_time = current_time
+        return cls.models_cache
+
+    def ensure_3d_tensor(self, tensor):
+        if tensor.dim() == 1:
+            return tensor.unsqueeze(0).unsqueeze(0)
+        elif tensor.dim() == 2:
+            return tensor.unsqueeze(0)
+        elif tensor.dim() > 3:
+            return tensor.squeeze().unsqueeze(0)
+        return tensor
+
+    def _load_audio_from_bytes(self, audio_bytes: bytes):
+        """Decode ElevenLabs MP3 response into torch waveform."""
+        if not audio_bytes:
+            logging.error("Empty audio response received.")
+            return None, None
+
+        try:
+            audio = AudioSegment.from_file(io.BytesIO(audio_bytes), format="mp3")
+        except Exception as e:
+            preview = audio_bytes[:80]
+            try:
+                preview_text = preview.decode("utf-8", errors="replace")
+            except Exception:
+                preview_text = str(preview)
+            logging.error(f"Error decoding MP3 audio: {e}. Preview: {preview_text}")
+            return None, None
+
+        sample_rate = audio.frame_rate
+        sample_width = audio.sample_width
+        channels = audio.channels
+
+        samples = np.array(audio.get_array_of_samples())
+        if channels > 1:
+            samples = samples.reshape(-1, channels).T  # channels x samples
+        else:
+            samples = samples.reshape(1, -1)
+
+        max_val = float(2 ** (8 * sample_width - 1))
+        waveform = torch.from_numpy(samples).unsqueeze(0).float() / max_val
+        waveform = self.ensure_3d_tensor(waveform)
+
+        return waveform, sample_rate
+
+    def generate_speech(self, text, voice, model, stability, similarity_boost, style, use_speaker_boost, input_text=None):
+        # Use input_text if provided, otherwise use the text from the textbox
+        final_text = input_text if input_text is not None else text
+
+        voice_id = self.voices_map.get(voice)
+        if not voice_id:
+            logging.error(f"Voice ID not found for selection '{voice}'. Refreshing voices cache.")
+            self.fetch_elevenlabs_voices()
+            voice_id = self.voices_map.get(voice)
+            if not voice_id:
+                return ({"waveform": torch.zeros(1, 1, 1).float(), "sample_rate": 44100},)
+
+        api_key = self._get_api_key()
+        if api_key is None:
+            return ({"waveform": torch.zeros(1, 1, 1).float(), "sample_rate": 44100},)
+
+        headers = {
+            "Accept": "audio/mpeg",
+            "xi-api-key": api_key,
+            "Content-Type": "application/json"
+        }
+
+        # Text-to-Speech
+        url = f"https://api.elevenlabs.io/v1/text-to-speech/{voice_id}"
+        payload = {
+            "text": final_text,
+            "model_id": model,
+            "voice_settings": {
+                "stability": stability,
+                "similarity_boost": similarity_boost,
+                "style": style,
+                "use_speaker_boost": use_speaker_boost
+            }
+        }
+        try:
+            response = requests.post(url, headers=headers, json=payload)
+        except requests.exceptions.RequestException as e:
+            logging.error(f"Error in text-to-speech: {str(e)}")
+            return ({"waveform": torch.zeros(1, 1, 1).float(), "sample_rate": 44100},)
+
+        if response.status_code == 200:
+            waveform, sample_rate = self._load_audio_from_bytes(response.content)
+            if waveform is None or sample_rate is None:
+                return ({"waveform": torch.zeros(1, 1, 1).float(), "sample_rate": 44100},)
+
+            # Ensure the waveform is 3D
+            if waveform.dim() != 3:
+                logging.info(f"Unexpected tensor dimension {waveform.dim()}. Reshaping to 3D.")
+                waveform = waveform.view(1, 1, -1)
+            
+            return ({"waveform": waveform, "sample_rate": sample_rate},)
+        else:
+            logging.error(f"API Error: {response.status_code} - {response.text}")
+            return ({"waveform": torch.zeros(1, 1, 1).float(), "sample_rate": 44100},)
+
+    @classmethod
+    def IS_CHANGED(cls, text, voice, model, stability, similarity_boost, style, use_speaker_boost, input_text=None):
+        return (text, voice, model, stability, similarity_boost, style, use_speaker_boost, input_text)
+
+# Node class mappings
+NODE_CLASS_MAPPINGS = {
+    "ElevenLabsNode": ElevenLabsNode
+}
+
+# Node display name mappings
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "ElevenLabsNode": "ElevenLabs TTS Node"
+}

requirements.txt

@@ -0,0 +1,4 @@
+requests
+numpy
+pydub
+torch