# _`FastKoko`_
[]()
[]()
[](https://huggingface.co/hexgrad/Kokoro-82M/tree/c3b0d86e2a980e027ef71c28819ea02e351c2667) [](https://huggingface.co/spaces/Remsky/Kokoro-TTS-Zero)
> Support for Kokoro-82M v1.0 coming very soon! Dev build on the `v0.1.5-integration` branch
Dockerized FastAPI wrapper for [Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M) text-to-speech model
- OpenAI-compatible Speech endpoint, with inline voice combination, and mapped naming/models for strict systems
- NVIDIA GPU accelerated or CPU inference (ONNX or Pytorch for either)
- very fast generation time
- ~35x-100x+ real time speed via 4060Ti+
- ~5x+ real time speed via M3 Pro CPU
- streaming support & tempfile generation, phoneme based dev endpoints
- (new) Integrated web UI on localhost:8880/web
- (new) Debug endpoints for monitoring threads, storage, and session pools
## Get Started
Quickest Start (docker run)
Pre built images are available to run, with arm/multi-arch support, and baked in models
Refer to the core/config.py file for a full list of variables which can be managed via the environment
```bash
docker run -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-cpu:v0.1.4 # CPU, or:
docker run --gpus all -p 8880:8880 ghcr.io/remsky/kokoro-fastapi-gpu:v0.1.4 #NVIDIA GPU
```
Quick Start (docker compose)
1. Install prerequisites, and start the service using Docker Compose (Full setup including UI):
- Install [Docker](https://www.docker.com/products/docker-desktop/)
- Clone the repository:
```bash
git clone https://github.com/remsky/Kokoro-FastAPI.git
cd Kokoro-FastAPI
cd docker/gpu # OR
# cd docker/cpu # Run this or the above
docker compose up --build
# if you are missing any models, run:
# python ../scripts/download_model.py --type pth # for GPU
# python ../scripts/download_model.py --type onnx # for CPU
```
```bash
Or directly via UV
./start-cpu.sh
./start-gpu.sh
```
Direct Run (via uv)
1. Install prerequisites ():
- Install [astral-uv](https://docs.astral.sh/uv/)
- Clone the repository:
```bash
git clone https://github.com/remsky/Kokoro-FastAPI.git
cd Kokoro-FastAPI
# if you are missing any models, run:
# python ../scripts/download_model.py --type pth # for GPU
# python ../scripts/download_model.py --type onnx # for CPU
```
Start directly via UV (with hot-reload)
```bash
./start-cpu.sh OR
./start-gpu.sh
```
Up and Running?
Run locally as an OpenAI-Compatible Speech Endpoint
```python
from openai import OpenAI
client = OpenAI(
base_url="http://localhost:8880/v1", api_key="not-needed"
)
with client.audio.speech.with_streaming_response.create(
model="kokoro",
voice="af_sky+af_bella", #single or multiple voicepack combo
input="Hello world!"
) as response:
response.stream_to_file("output.mp3")
```
- The API will be available at http://localhost:8880
- API Documentation: http://localhost:8880/docs
- Web Interface: http://localhost:8880/web
- Gradio UI (deprecating) can be accessed at http://localhost:7860 if enabled in docker compose file (it is a separate image!)
## Features
OpenAI-Compatible Speech Endpoint
```python
# Using OpenAI's Python library
from openai import OpenAI
client = OpenAI(base_url="http://localhost:8880/v1", api_key="not-needed")
response = client.audio.speech.create(
model="kokoro",
voice="af_bella+af_sky", # see /api/src/core/openai_mappings.json to customize
input="Hello world!",
response_format="mp3"
)
response.stream_to_file("output.mp3")
```
Or Via Requests:
```python
import requests
response = requests.get("http://localhost:8880/v1/audio/voices")
voices = response.json()["voices"]
# Generate audio
response = requests.post(
"http://localhost:8880/v1/audio/speech",
json={
"model": "kokoro",
"input": "Hello world!",
"voice": "af_bella",
"response_format": "mp3", # Supported: mp3, wav, opus, flac
"speed": 1.0
}
)
# Save audio
with open("output.mp3", "wb") as f:
f.write(response.content)
```
Quick tests (run from another terminal):
```bash
python examples/assorted_checks/test_openai/test_openai_tts.py # Test OpenAI Compatibility
python examples/assorted_checks/test_voices/test_all_voices.py # Test all available voices
```
Voice Combination
- Averages model weights of any existing voicepacks
- Saves generated voicepacks for future use
- (new) Available through any endpoint, simply concatenate desired packs with "+"
Combine voices and generate audio:
```python
import requests
response = requests.get("http://localhost:8880/v1/audio/voices")
voices = response.json()["voices"]
# Create combined voice (saves locally on server)
response = requests.post(
"http://localhost:8880/v1/audio/voices/combine",
json=[voices[0], voices[1]]
)
combined_voice = response.json()["voice"]
# Generate audio with combined voice (or, simply pass multiple directly with `+` )
response = requests.post(
"http://localhost:8880/v1/audio/speech",
json={
"input": "Hello world!",
"voice": combined_voice, # or skip the above step with f"{voices[0]}+{voices[1]}"
"response_format": "mp3"
}
)
```
Multiple Output Audio Formats
- mp3
- wav
- opus
- flac
- aac
- pcm
Gradio Web Utility
Access the interactive web UI at http://localhost:7860 after starting the service. Features include:
- Voice/format/speed selection
- Audio playback and download
- Text file or direct input
If you only want the API, just comment out everything in the docker-compose.yml under and including `gradio-ui`
Currently, voices created via the API are accessible here, but voice combination/creation has not yet been added
Running the UI Docker Service [deprecating]
- If you only want to run the Gradio web interface separately and connect it to an existing API service:
```bash
docker run -p 7860:7860 \
-e API_HOST= \
-e API_PORT=8880 \
```
- Replace `` with:
- `kokoro-tts` if the UI container is running in the same Docker Compose setup.
- `localhost` if the API is running on your local machine.
### Disabling Local Saving
You can disable local saving of audio files and hide the file view in the UI by setting the `DISABLE_LOCAL_SAVING` environment variable to `true`. This is useful when running the service on a server where you don't want to store generated audio files locally.
When using Docker Compose:
```yaml
environment:
- DISABLE_LOCAL_SAVING=true
```
When running the Docker image directly:
```bash
docker run -p 7860:7860 -e DISABLE_LOCAL_SAVING=true ghcr.io/remsky/kokoro-fastapi-ui:v0.1.4
```
Streaming Support
```python
# OpenAI-compatible streaming
from openai import OpenAI
client = OpenAI(
base_url="http://localhost:8880/v1", api_key="not-needed")
# Stream to file
with client.audio.speech.with_streaming_response.create(
model="kokoro",
voice="af_bella",
input="Hello world!"
) as response:
response.stream_to_file("output.mp3")
# Stream to speakers (requires PyAudio)
import pyaudio
player = pyaudio.PyAudio().open(
format=pyaudio.paInt16,
channels=1,
rate=24000,
output=True
)
with client.audio.speech.with_streaming_response.create(
model="kokoro",
voice="af_bella",
response_format="pcm",
input="Hello world!"
) as response:
for chunk in response.iter_bytes(chunk_size=1024):
player.write(chunk)
```
Or via requests:
```python
import requests
response = requests.post(
"http://localhost:8880/v1/audio/speech",
json={
"input": "Hello world!",
"voice": "af_bella",
"response_format": "pcm"
},
stream=True
)
for chunk in response.iter_content(chunk_size=1024):
if chunk:
# Process streaming chunks
pass
```
Key Streaming Metrics:
- First token latency @ chunksize
- ~300ms (GPU) @ 400
- ~3500ms (CPU) @ 200 (older i7)
- ~<1s (CPU) @ 200 (M3 Pro)
- Adjustable chunking settings for real-time playback
*Note: Artifacts in intonation can increase with smaller chunks*
## Processing Details
Performance Benchmarks
Benchmarking was performed on generation via the local API using text lengths up to feature-length books (~1.5 hours output), measuring processing time and realtime factor. Tests were run on:
- Windows 11 Home w/ WSL2
- NVIDIA 4060Ti 16gb GPU @ CUDA 12.1
- 11th Gen i7-11700 @ 2.5GHz
- 64gb RAM
- WAV native output
- H.G. Wells - The Time Machine (full text)
Key Performance Metrics:
- Realtime Speed: Ranges between 35x-100x (generation time to output audio length)
- Average Processing Rate: 137.67 tokens/second (cl100k_base)
GPU Vs. CPU
```bash
# GPU: Requires NVIDIA GPU with CUDA 12.1 support (~35x-100x realtime speed)
docker compose up --build
# CPU: ONNX optimized inference (~5x+ realtime speed on M3 Pro)
docker compose -f docker-compose.cpu.yml up --build
```
*Note: Overall speed may have reduced somewhat with the structural changes to accomodate streaming. Looking into it*
Natural Boundary Detection
- Automatically splits and stitches at sentence boundaries
- Helps to reduce artifacts and allow long form processing as the base model is only currently configured for approximately 30s output
Phoneme & Token Routes
Convert text to phonemes and/or generate audio directly from phonemes:
```python
import requests
def get_phonemes(text: str, language: str = "a"):
"""Get phonemes and tokens for input text"""
response = requests.post(
"http://localhost:8880/dev/phonemize",
json={"text": text, "language": language} # "a" for American English
)
response.raise_for_status()
result = response.json()
return result["phonemes"], result["tokens"]
def generate_audio_from_phonemes(phonemes: str, voice: str = "af_bella"):
"""Generate audio from phonemes"""
response = requests.post(
"http://localhost:8880/dev/generate_from_phonemes",
json={"phonemes": phonemes, "voice": voice},
headers={"Accept": "audio/wav"}
)
if response.status_code != 200:
print(f"Error: {response.text}")
return None
return response.content
# Example usage
text = "Hello world!"
try:
# Convert text to phonemes
phonemes, tokens = get_phonemes(text)
print(f"Phonemes: {phonemes}") # e.g. ðɪs ɪz ˈoʊnli ɐ tˈɛst
print(f"Tokens: {tokens}") # Token IDs including start/end tokens
# Generate and save audio
if audio_bytes := generate_audio_from_phonemes(phonemes):
with open("speech.wav", "wb") as f:
f.write(audio_bytes)
print(f"Generated {len(audio_bytes)} bytes of audio")
except Exception as e:
print(f"Error: {e}")
```
See `examples/phoneme_examples/generate_phonemes.py` for a sample script.
Debug Endpoints
Monitor system state and resource usage with these endpoints:
- `/debug/threads` - Get thread information and stack traces
- `/debug/storage` - Monitor temp file and output directory usage
- `/debug/system` - Get system information (CPU, memory, GPU)
- `/debug/session_pools` - View ONNX session and CUDA stream status
Useful for debugging resource exhaustion or performance issues.
## Known Issues
Versioning & Development
I'm doing what I can to keep things stable, but we are on an early and rapid set of build cycles here.
If you run into trouble, you may have to roll back a version on the release tags if something comes up, or build up from source and/or troubleshoot + submit a PR. Will leave the branch up here for the last known stable points:
`v0.0.5post1`
Free and open source is a community effort, and I love working on this project, though there's only really so many hours in a day. If you'd like to support the work, feel free to open a PR, buy me a coffee, or report any bugs/features/etc you find during use.
Linux GPU Permissions
Some Linux users may encounter GPU permission issues when running as non-root.
Can't guarantee anything, but here are some common solutions, consider your security requirements carefully
### Option 1: Container Groups (Likely the best option)
```yaml
services:
kokoro-tts:
# ... existing config ...
group_add:
- "video"
- "render"
```
### Option 2: Host System Groups
```yaml
services:
kokoro-tts:
# ... existing config ...
user: "${UID}:${GID}"
group_add:
- "video"
```
Note: May require adding host user to groups: `sudo usermod -aG docker,video $USER` and system restart.
### Option 3: Device Permissions (Use with caution)
```yaml
services:
kokoro-tts:
# ... existing config ...
devices:
- /dev/nvidia0:/dev/nvidia0
- /dev/nvidiactl:/dev/nvidiactl
- /dev/nvidia-uvm:/dev/nvidia-uvm
```
⚠️ Warning: Reduces system security. Use only in development environments.
Prerequisites: NVIDIA GPU, drivers, and container toolkit must be properly configured.
Visit [NVIDIA Container Toolkit installation](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html) for more detailed information
## Model and License
Model
This API uses the [Kokoro-82M](https://huggingface.co/hexgrad/Kokoro-82M) model from HuggingFace.
Visit the model page for more details about training, architecture, and capabilities. I have no affiliation with any of their work, and produced this wrapper for ease of use and personal projects.
License
This project is licensed under the Apache License 2.0 - see below for details:
- The Kokoro model weights are licensed under Apache 2.0 (see [model page](https://huggingface.co/hexgrad/Kokoro-82M))
- The FastAPI wrapper code in this repository is licensed under Apache 2.0 to match
- The inference code adapted from StyleTTS2 is MIT licensed
The full Apache 2.0 license text can be found at: https://www.apache.org/licenses/LICENSE-2.0
