Whisper Example with Hailo Acceleration!

Community Projects
1

:speaking_head: New Release: Whisper Example with Hailo Acceleration!

We’re excited to announce the release of our Whisper example – showcasing how to accelerate OpenAI’s powerful speech-to-text model on Hailo hardware.

:brain: Want to get started?
Check out the example here:
:backhand_index_pointing_right: Whisper Speech Recognition Example on GitHub

:rocket: Community Challenge: Make Whisper Work for Everyone

To celebrate this release, we’re launching a community contest!
Your mission: Integrate our Whisper acceleration into an existing third-party solution and make it accessible to the wider community.

:bullseye: Suggested Targets:

  • :white_check_mark: Integrate with the Wyoming Protocol
  • :white_check_mark: Plug into open-source voice assistant stacks
  • :white_check_mark: Contribute to Home Assistant, Rhasspy, or other community-driven tools

:puzzle_piece: What We’re Looking For:

  • :unlocked: Open-source integration (public GitHub repo)
  • :handshake: Easy to use and well-documented for the community
  • :high_voltage: Uses Hailo’s accelerated Whisper model

:trophy: Prizes & Recognition:

Top projects will be featured on our Community and promoted across our channels.
To participate, share your integration in the Community Forum as a Community Project, and don’t forget to add the Whisper tag!

:date: Submission deadline: July 1st
:sports_medal: The top 2 most-liked projects will each receive:

  • One 26 TOPS AI HAT+
  • One $100 Amazon Gift Card

Let’s make speech AI accessible for everyone — together. :speech_balloon:

6 Likes
3

Fantastic work. Also, thank you for keeping the community updated as you developed this solution. Looking forward to integrating with my project!

1 Like
4

This is fantastic, thanks! Is it possible to use our own fine-tuned whisper-tiny model? I am assuming there would need to be a fair amount of conversion to get it in the right shape for this pipeline?

5

Hi @Andrew_Evans we will release the compilation scripts soon. The compilation requires a feature that we added in a DFC version which is not yet supported on the Raspberry Pi.
We will keep the community updated.