Skip to content

Commit 4124f27

Browse files
tszumowskitszumowski
authored
Add GCP Text-to-Speech and Cloud Run (#1)
* add gcp tts interface * add TODOs to readme * add optional TTS and adjust autoplay * more updates * add precommit and flake8 * parameterize more * parameterize more * formatting * config update * add dockerfile * update readme * docker updates * more docker updates, now tested * more docker updates, now tested * add cloud run tested
1 parent 33fde8e commit 4124f27

8 files changed

+399
-35
lines changed

‎.flake8

+18
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,18 @@
1+
[flake8]
2+
# Some sane defaults for the code style checker flake8
3+
exclude =
4+
.tox
5+
build
6+
dist
7+
.eggs
8+
docs/conf.py
9+
max-line-length = 88
10+
ignore =
11+
# Whitespace before ':'
12+
E203
13+
# Whitepsace at end of line
14+
W291
15+
# Line break before logical
16+
W503
17+
# Missing f-string placeholders
18+
F541

‎.gitignore

+8
Original file line numberDiff line numberDiff line change
@@ -127,3 +127,11 @@ dmypy.json
127127

128128
# Pyre type checker
129129
.pyre/
130+
131+
# Other
132+
**/*.jpg
133+
**/*.jpeg
134+
**/*.mp3
135+
**/*.wav
136+
**/*.json
137+
**/*.txt

‎.pre-commit-config.yaml

+31
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,31 @@
1+
---
2+
fail_fast: true
3+
repos:
4+
- repo: https://github.com/pre-commit/pre-commit-hooks
5+
rev: v2.1.0
6+
hooks:
7+
- id: check-executables-have-shebangs
8+
- id: check-json
9+
- id: pretty-format-json
10+
args: ["--autofix"]
11+
- id: check-merge-conflict
12+
- id: debug-statements
13+
- id: detect-private-key
14+
- id: forbid-new-submodules
15+
- id: trailing-whitespace
16+
- id: requirements-txt-fixer
17+
- repo: https://github.com/adrienverge/yamllint
18+
rev: v1.14.0
19+
hooks:
20+
- id: yamllint
21+
args: ['-d {rules: {line-length: disable}}', '-s']
22+
- repo: https://github.com/ambv/black
23+
rev: 22.3.0
24+
hooks:
25+
- id: black
26+
language_version: python3
27+
- repo: https://github.com/pycqa/flake8
28+
rev: 3.8.3
29+
hooks:
30+
- id: flake8
31+
args: ["--config=.flake8"]

‎Dockerfile

+23
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,23 @@
1+
# Use the official Python image as the parent image
2+
FROM python:3.8-slim-buster
3+
4+
# Set the working directory to /app
5+
WORKDIR /app
6+
7+
# Install FFmpeg and other system dependencies
8+
RUN apt-get update && apt-get install -y --no-install-recommends \
9+
ffmpeg \
10+
&& rm -rf /var/lib/apt/lists/*
11+
12+
# Install the required dependencies
13+
COPY requirements.txt ./
14+
RUN pip install --no-cache-dir -r requirements.txt
15+
16+
# Copy the necessary files to the Docker image
17+
COPY storyteller.py config.py ./
18+
19+
# Expose port 7860
20+
EXPOSE 7860
21+
22+
# Set the default command to execute the `storyteller.py` script
23+
CMD ["python", "storyteller.py"]

‎README.md

+95-9
Original file line numberDiff line numberDiff line change
@@ -2,21 +2,107 @@
22

33
---
44

5-
TODO: Document. Medium?
6-
Make note of pricing.
5+
This is a Gradio UI application that takes in a request for a story from the microphone
6+
and speaks an interactive Choose-Your-Own-Adventure style children's story. It leverages:
7+
8+
- [OpenAI Whisper](https://openai.com/research/whisper): to transcribe user audio input request
9+
- [OpenAI ChatGPT (3.5-turbo)](https://platform.openai.com/docs/models/gpt-3-5):
10+
to generate a story chapter given the user's inputs
11+
- (Optional) [Google Cloud Text-to-Speech](https://cloud.google.com/text-to-speech/):
12+
to use realistic voices when telling the story.
13+
14+
## Pricing
15+
16+
**WARNING: This application uses paid API services. Create quotas and watch your usage.**
17+
18+
At the time of writing, the pricing is as follows:
19+
20+
- [whisper](https://openai.com/pricing): $0.006 / minute (rounded to the nearest second)
21+
- [gpt-3.5-turbo](https://openai.com/pricing): $0.002 / 1K tokens
22+
- [Google Text-to-Speech](https://cloud.google.com/text-to-speech/pricing):
23+
- 0 to 1 million bytes free per month
24+
- $0.000016 USD per byte ($16.00 USD per 1 million bytes)
25+
26+
Check the links as these can change often. But at the time of writing it costs less
27+
than one USD for light use.
28+
29+
Both OpenAI and Google offer free credits for new users.
730

831
## Setup
932

10-
1. Get an OpenAI API key.
33+
Note there are two ways to speak the story: Mac or GCP Text-to-Speech. If using a Mac,
34+
the Mac `say` command is used and that's the easiest/fastest route to running this.
35+
It uses the System voice set up in the Accessibility settings.
36+
However, if not on a Mac or if you prefer a more realistic voice, the GCP Text-to-Speech may be used.
37+
This requires you having (a) a GCP project, (b) the TTS API enabled, and (c) your account authenticated
38+
in gcloud (or GOOGLE_APPLICATION_CREDENTIALS environment variable set).
39+
40+
This application has only been tested on a Macbook.
41+
42+
1. Sign up at OpenAI and acquire an [OpenAI API key](https://platform.openai.com/account/api-keys).
1143
1. Add to environment variable with: `export OPENAI_API_KEY="sk-xxxxxxxxxxxxxxx`"
1244
1. Create virtual environment
13-
1. `pip install -r requirements.txt`
14-
1. Brew install `ffmpeg`: `brew install ffmpeg`
15-
1. Update config in `config.py` as desired
45+
1. Run `pip install -r requirements.txt`
46+
1. If on Mac, brew install `ffmpeg`: `brew install ffmpeg`
47+
48+
- Linux may need to install also but untested.
49+
50+
1. Review and update config in `config.py` as desired
51+
1. If using GCP TTS
52+
1. set in `config.py`: `SPEECH_METHOD = SpeechMethod.GCP`
53+
1. Navigate to the [Google API page](https://console.cloud.google.com/apis/api/texttospeech.googleapis.com/) and enable the API
54+
1. Confirm you are authenticated in gcloud and your account has access to that API.
1655
1. Run with: `python storyteller.py`
1756
1. Navigate to `http://127.0.0.1:7860/` and have fun!
1857

19-
## TODO
58+
## Running as Docker Container
59+
60+
Replace `<service-name>` with a name of your choice.
61+
62+
1. Build Docker image: `docker build -t <image-name> .`
63+
1. Run locally with something similar to:
64+
65+
```
66+
docker run -it --rm \
67+
-e GOOGLE_APPLICATION_CREDENTIALS=/tmp/creds.json \
68+
-v ${HOME}/.config/gcloud/application_default_credentials.json:/tmp/creds.json \
69+
-e OPENAI_API_KEY=<openai-api-key> \
70+
-p <port>:7860 \
71+
audio-storyteller \
72+
python storyteller.py \
73+
--address=0.0.0.0 \
74+
--port=7860 \
75+
--user=<username> \
76+
--password=<password>
77+
```
78+
79+
Fill in: `<openai-api-key>, <port>, and optional <username>:<password>.
80+
Then once running, navigate on a browser to `127.0.0.1:<port>` and fill in the
81+
optional username:password you provided.
82+
83+
## Deploying to Google Cloud Run
84+
85+
1. Follow the directions above to create a local docker image.
86+
1. Tag and push (Note: Follow [these directions](https://cloud.google.com/container-registry/docs/advanced-authentication) to authenticate)
87+
```
88+
docker tag <image-name> gcr.io/<project-id>/<image-name>
89+
docker push gcr.io/<project-id>/<image-name>
90+
```
91+
1. Create a service account on your GCP project IAM page named: `audio-storytelling-bot@<project-id>.iam.gserviceaccount.com`
92+
1. Deploy with the following command, setting anything in `<>` appropriately:
93+
94+
```
95+
gcloud run deploy audio-storytelling-bot \
96+
--image gcr.io/<project-id>/<image-name> \
97+
--platform managed \
98+
--service-account=audio-storytelling-bot@<project-id>.iam.gserviceaccount.com \
99+
--set-env-vars=OPENAI_API_KEY=<openai-key-string> \
100+
--no-allow-unauthenticated \
101+
--port=7860 \
102+
--cpu=1 \
103+
--memory=512Mi \
104+
--min-instances=0 \
105+
--max-instances=1
106+
```
20107

21-
- [ ] Fix the audio thread error that pops up
22-
- [ ] Document
108+
Cloud Run will automatically scale the number of instances based on the incoming traffic. You can access the deployed Gradio application via the URL provided by the Cloud Run service.

‎config.py

+38
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,41 @@
1+
from enum import Enum
12
import time
23

4+
"""
5+
Speech method
6+
None: No speech
7+
"gcp": Google Cloud Platform Text-to-Speech API
8+
"mac": Mac OS X say command
9+
10+
Note: For GCP, you must be authenticated with the gcloud CLI or set the
11+
GOOGLE_APPLICATION_CREDENTIALS environment variable
12+
"""
13+
14+
15+
# Define the class enum
16+
class SpeechMethod(Enum):
17+
NONE = 1
18+
GCP = 2
19+
MAC = 3
20+
21+
22+
# Set the method here
23+
SPEECH_METHOD = SpeechMethod.GCP
24+
25+
26+
"""
27+
Other configuration
28+
"""
329
RESOLUTION = "512x512" # One of 256x256, 512x512, 1024x1024
430
PROMPT_MAX_LEN = 1000 # Max length of prompt for DALL-E
531
IMAGE_PATH = "generated_image.jpg" # path to save generated image
632
TRANSCRIPT_PATH = f"transcript-{int(time.time())}.txt"
33+
GENERATED_SPEECH_PATH = "generated_speech.mp3"
34+
TTS_SPEECH_DELAY = 5.0 # seconds to wait before playing generated speech
35+
36+
# Voice for GCP Text-to-Speech API
37+
# Samples: https://cloud.google.com/text-to-speech/docs/voices
38+
TTS_VOICE = "en-GB-Neural2-C"
739

840
"""
941
Example Prompts
@@ -33,3 +65,9 @@
3365
of each chapter first pause for a moment. Then ask the reader a single question
3466
that chooses the path for their next chapter in their story.
3567
"""
68+
69+
"""
70+
DERIVED CONFIG
71+
"""
72+
# Derive only xx-xx from TTS_VOICE
73+
TTS_VOICE_LANGUAGE_CODE = "-".join(TTS_VOICE.split("-")[0:2])

‎requirements.txt

+5-1
Original file line numberDiff line numberDiff line change
@@ -1,2 +1,6 @@
1+
black
2+
flake8
3+
google-cloud-texttospeech
4+
gradio
15
openai
2-
gradio
6+
pre-commit

0 commit comments

Comments
 (0)