It is a Sphinx's builder to generate MP3 files from reST files by using Amazon Polly.
- It can generate SSML format files.
- It kicks Amazon Polly and get MP3 fragments.
- Finally it kicks ffmpeg to concat MP3 fragments into one MP3 file.
Amazon Polly API has text size limitation (1000 chracters without SSML tags). This builder generates small valid size SSML files for Polly. It caches MP3 fragments to reduce API access (it keeps hashes and reuse it).
- boto3, awscli (see last note at : http://docs.aws.amazon.com/polly/latest/dg/examples-python.html)
- AWS IAM user (that has Polly read access right. Few regions support Polly.)
- ffmpeg
Getting Started with Amazon Polly is a good entry point to setup.
pip install sphinxcontrib-ssmlbuilderAdd setting like this:
extensions = ['sphinxcontrib.ssmlbuilder']
# you should register the following user name
# via aws configure.
ssml_polly_aws_profile = "polly_user"
ssml_polly_apply_docnames = "*"
# for Japanese
ssml_language = 'ja-JP'
ssml_polly_aws_voiceid = 'Mizuki'| Config Key | Default Value | Description | Available values |
|---|---|---|---|
ssml_language |
"en-US" |
SSML speaking language option. | AWS Doc |
ssml_skip_block |
{'comment': True, 'table': True, 'codeblock': True} |
If the value is True, the content inside the blocks are omitted from resulting SSML. | |
ssml_break_around_section_title |
[2000, 1600, 1000, 1000, 1000, 1000] |
Break time (millisecond) around section title. | |
ssml_emphasis_section_title |
['none', 'none', 'none', 'none', 'none', 'none'] |
Emphasis level of section titles | "strong", "moderate", "none", "reduced" |
ssml_paragraph_speed |
"default" |
Speech speed | "x-slow", "slow", "medium", "fast", "x-fast", "default" |
ssml_break_after_paragraph |
1000 |
Break time after paragraph (millisecond). | |
ssml_polly_audio_output_folder |
polly |
Output folder for resulting files. It uses outside of build folder to avoid make clean to save money. |
|
ssml_polly_aws_profile |
'' |
AWS Profile name that is configured by aws configure command. |
|
ssml_polly_aws_voice_id |
"Joanna" |
Voice ID to use for the Polly synthesis. | AWS Doc |
ssml_polly_apply_docnames |
'' |
Docname pattern you want to adopt Polly.
If you want to convert only "readme.rst" on top folder, set "readme".
"*" matches all document.
Default value doesn't match with any files. |
BSD License
Yoshiki Shibukawa
https://github.com/shibukawa/sphinxcontrib-ssmlbuilder
- 0.2.0 Jan 29 2017
- Exec AWS API parallelly.
- 0.1.3 Jan 24 2017
- Remove extra doublequote in MP3 metadata
- 0.1.2 Jan 24 2017
- Change default speed to
default - Remove duplicated breaks after paragraphs
- Add metadata to resulting MP3 (title, author, track, year, genre)
- Change default speed to
- 0.1.1 Jan 23 2017
- Add skipping block option
- 0.1 Jan 23 2017
- Initial Version