Added serverless API

README.md

@@ -73,6 +73,10 @@ It is possible to make minor changes to the generated speech.
 
 In addition, it is required to use `--rate=-50%` instead of `--rate -50%` (note the lack of an equal sign) otherwise the `-50%` would be interpreted as just another argument.
 
+### Deploying to Serverless API
+
+See the [examples/serverless-api](examples/serverless-api) folder for more information on how to deploy to [Cerebrium](https://www.cerebrium.ai).
+
 ### Note on the `edge-playback` command
 
 `edge-playback` is just a wrapper around `edge-tts` that plays back the generated speech. It takes the same arguments as the `edge-tts` option.

examples/serverless-api/README.md

@@ -0,0 +1,33 @@
+# Serverless Edge TTS API
+
+This project demonstrates how to run Edge TTS as a serverless API using [Cerebrium](https://www.cerebrium.ai)
+
+## Overview
+
+The `main.py` file contains a function `run` that takes a text input and an optional voice parameter to generate audio and subtitles using Edge TTS.
+
+## Installation
+
+1. pip install cerebrium
+2. cerebrium login
+3. Make sure you are in the serverless-api folder and run ```cerebrium deploy```
+
+## Usage
+
+Once deployed, you should be able to make a curl request similar to:
+```
+curl --location 'https://api.cortex.cerebrium.ai/v4/p-xxxxxx/serverless-api/run' \
+--header 'Authorization: Bearer <AUTH_TOKEN>' \
+--header 'Content-Type: application/json' \
+--data '{"text": "Tell me something"}'
+```
+
+The `run` function takes two parameters:
+
+- `text` (str): The text to be converted to speech
+- `voice` (str, optional): The voice to use for TTS (default: "en-GB-SoniaNeural")
+
+It returns a dictionary containing:
+
+- `audio_data`: The generated audio as a base64-encoded string
+- `subtitles`: The generated subtitles in WebVTT format

examples/serverless-api/cerebrium.toml

@@ -0,0 +1,19 @@
+[cerebrium.deployment]
+name = "serverless-api"
+python_version = "3.11"
+docker_base_image_url = "debian:bookworm-slim"
+include = "[./*, main.py, cerebrium.toml]"
+exclude = "[.*]"
+
+[cerebrium.hardware]
+cpu = 2
+memory = 12.0
+compute = "CPU"
+
+[cerebrium.scaling]
+min_replicas = 0
+max_replicas = 5
+cooldown = 30
+
+[cerebrium.dependencies.pip]
+"edge-tts" = "latest"

examples/serverless-api/main.py

@@ -0,0 +1,21 @@
+
+import edge_tts
+
+async def run(text: str, voice: str = "en-GB-SoniaNeural"): 
+
+ communicate = edge_tts.Communicate(text, voice)
+ submaker = edge_tts.SubMaker()
+ audio_data = bytearray()
+ subtitles = ""
+
+ async for chunk in communicate.stream():
+ if chunk["type"] == "audio":
+ audio_data.extend(chunk["data"])
+ elif chunk["type"] == "WordBoundary":
+ submaker.create_sub((chunk["offset"], chunk["duration"]), chunk["text"])
+
+ subtitles = submaker.generate_subs()
+ return {
+ "audio_data": audio_data.decode("latin-1"),
+ "subtitles": subtitles
+ }