Hey Product Hunt, Gustav here from Epilude. Epilude is the voice productivity suite for your Mac we wish Apple would build, with dictation, meeting recording and now, text-to-speech with voices you'll actually like listening to with Narrator.
Today we're expanding Epilude with Narrator, a completely private text-to-speech tool that narrates any test on your Mac without any of your data leaving your machine.
Here's how it works: - Select any text in any app on your Mac, or point it at a document or public URL (like that article you're interested in but never found time to sit down and read). Narrator will process the text locally using our local voice model EP-1 voice, and turn it into a narrated sound clip without any data leaving your Mac.
- As your narration starts, a small player appears and reads your text aloud, highlighting each word as it reads to you. We've found this often helps a lot with comprehension, as you can both listen to the text and read along at the same time.
- You can send your narration to your Epilude account on web, so that you can access it from anywhere and listen when you're on the go. This also lets you share your narrations with anyone you want.
You can try Narrator completely for free. We're really excited for you to try it out. Happy listening!
Hey Product Hunt - Mark here, the other half of Epilude.
I've had the same dozen articles open in tabs for months, fully intending to read them. These days I select one before I leave the desk, Narrator turns it into audio on the Mac, and I listen to it while walking the dog. I've got through more of them in the last few weeks than in the previous year.
Yogi has now sat through three longreads on Postgres indexing. Not sure she's convinced. Would be good to hear what you make of it.
the local processing is the right call for this specific use case, text-to-speech doesn't need a round trip to be good, so keeping it on-device is just strictly better and you get privacy for free instead of as a tradeoff. question on the on-device EP-1 voice though, how does it hold up over a genuinely long article, 20+ minutes, versus the cloud voice option. on-device models for speech tend to be the first place quality gets sacrificed for size
@galdayan ElevenLabs V4 (the cloud voice model we use) is noticeably ahead, but if you haven't tested a local voice model like EP-1 voice lately I think you'll be surprised by how good they are! Quality doesn't drop on longer articles, they just take slightly longer to process.
@gustav_ek good to hear, I'll give EP-1 a real test run then. "slightly longer" is the part I'd want a number on though, since for a 20-minute article the difference between 1.2x and 3x realtime decides whether this replaces a cloud call or just becomes a thing I kick off and walk away from. is there a rough multiplier you'd quote for an M-series chip, or does it vary too much by article length to say
Replies
Hey Product Hunt, Gustav here from Epilude. Epilude is the voice productivity suite for your Mac we wish Apple would build, with dictation, meeting recording and now, text-to-speech with voices you'll actually like listening to with Narrator.
Today we're expanding Epilude with Narrator, a completely private text-to-speech tool that narrates any test on your Mac without any of your data leaving your machine.
Here's how it works:
- Select any text in any app on your Mac, or point it at a document or public URL (like that article you're interested in but never found time to sit down and read). Narrator will process the text locally using our local voice model EP-1 voice, and turn it into a narrated sound clip without any data leaving your Mac.
- As your narration starts, a small player appears and reads your text aloud, highlighting each word as it reads to you. We've found this often helps a lot with comprehension, as you can both listen to the text and read along at the same time.
- You can send your narration to your Epilude account on web, so that you can access it from anywhere and listen when you're on the go. This also lets you share your narrations with anyone you want.
You can try Narrator completely for free. We're really excited for you to try it out. Happy listening!
Hey Product Hunt - Mark here, the other half of Epilude.
I've had the same dozen articles open in tabs for months, fully intending to read them. These days I select one before I leave the desk, Narrator turns it into audio on the Mac, and I listen to it while walking the dog. I've got through more of them in the last few weeks than in the previous year.
Yogi has now sat through three longreads on Postgres indexing. Not sure she's convinced. Would be good to hear what you make of it.
P.S. of course this was written with Epilude 😉
the local processing is the right call for this specific use case, text-to-speech doesn't need a round trip to be good, so keeping it on-device is just strictly better and you get privacy for free instead of as a tradeoff. question on the on-device EP-1 voice though, how does it hold up over a genuinely long article, 20+ minutes, versus the cloud voice option. on-device models for speech tend to be the first place quality gets sacrificed for size
@galdayan ElevenLabs V4 (the cloud voice model we use) is noticeably ahead, but if you haven't tested a local voice model like EP-1 voice lately I think you'll be surprised by how good they are! Quality doesn't drop on longer articles, they just take slightly longer to process.
@gustav_ek good to hear, I'll give EP-1 a real test run then. "slightly longer" is the part I'd want a number on though, since for a 20-minute article the difference between 1.2x and 3x realtime decides whether this replaces a cloud call or just becomes a thing I kick off and walk away from. is there a rough multiplier you'd quote for an M-series chip, or does it vary too much by article length to say