Mozilla has created a new project called "Common Voice" with the goal of collecting 10,000 hours of speech.  They have created a very nice web app to collect and validate submitted speech.  Their goal is to collect 10,000 of speech.  From their website:

What is Common Voice?

[...] Common Voice is a project to make voice recognition technology easily accessible to everyone. People donate their voices to a massive database that will let anyone quickly and easily train voice-enabled apps. All voice data will be available to developers.

[...]

When will the dataset be available?

Mozilla aims to begin to capture voices in June and release the open source database later in 2017.

All speech submitted will be released under the CC-0 license (public domain).  They are starting with English, and acknowledge the need to add more languages.