💎 Aurora_lily 💎 AI & Social Media Creator 📈 Turning Ideas Into Engagement 🤝 Open For Collaborations 📩 [email protected]
Camden Town, London
Joined March 2015
- Tweets45.6K
- Following27
- Followers32.5K
- Likes88.3K
Pinned Tweet
🚨𝗕𝗥𝗘𝗔𝗞𝗜𝗡𝗚: GOODBYE CHATGPT forever.
Claude just collapsed every tool you were using into one single 120-second system completely free.
7 prompts go from completely across 10 tools to completely unstoppable with just one: 👇
Aurora_lily retweeted
POV: Your product ad got a little too ambitious and decided to break out of the frame. 😂 And while you're making your next creative masterpiece, don't miss 60% off CapCut Pro or 40% off CapCut Ultra. Black Friday deals end Oct 30, so grab yours before they're gone!
Okay, this is the kind of visual that makes people look twice. You've got the giant portrait, someone is reaching in, the product reveal, and those close-up shots of the product itself. That's a lot of footage and visual elements to keep track of! CapCut PC's Multi-Track Editing lets you organise your clips, overlays and other assets on separate tracks, so you can adjust each layer without having to rearrange everything else.
CapCut Sale: tinyurl.com/2v23rn22
#capcutpc #capcutmarketing #capcutblackfriday #capcutsale
🤝 Paid partnership
Aurora_lily retweeted
I want human-like speech research to advance. I also want people to know when they are interacting with AI. Those positions are compatible.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
How should a score above 50% be interpreted? Systematic misidentification is interesting, but it isn't the same thing as random guessing.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
A single engaging audio pair can explain the test. It can't substitute for the full evaluation.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
@voicearena_ai 's question isn't just how close are we. It's what exactly are we trying to get close to, and why?
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
A strong Google result doesn't settle whether sounding human is what users need. Capability and product choice remain different decisions.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
The conversation should outlast the Google result. What does the next model teach us about human identification that this one doesn't?
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
A ranking can tell you who leads the tested group. Without a human comparison, it doesn't directly answer whether listeners can distinguish them from people.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
Creating a complete Al film from a script sounds complicated, but DeeVid Al is making the workflow much simpler.
Story development, video generation, voiceovers, subtitles, and editing in one place.
deevid.ai/app/ai-long-video
Turn Your Story into a Complete Video. 🎬
An all-in-one long-video workflow with no editing required.
Create videos up to 30 minutes with:
Script to Video
Continuous Storytelling
Consistent Characters
Multi-modal Input
AI Production Agent
From story development and storyboarding to scene creation, voiceovers, and subtitles — everything in one workflow.
Your story deserves to be told in full.
Try Deevid AI Long Video
#DeevidAI #AILongVideo #AIVideo #AIFilmmaking
Aurora_lily retweeted
The real potential of Al video isn't just making short clips. It's creating stories with a beginning, middle, and ending.
DeeVid Al is exploring exactly that.
deevid.ai/app/ai-long-video
Turn Your Story into a Complete Video. 🎬
An all-in-one long-video workflow with no editing required.
Create videos up to 30 minutes with:
Script to Video
Continuous Storytelling
Consistent Characters
Multi-modal Input
AI Production Agent
From story development and storyboarding to scene creation, voiceovers, and subtitles — everything in one workflow.
Your story deserves to be told in full.
Try Deevid AI Long Video
#DeevidAI #AILongVideo #AIVideo #AIFilmmaking
Aurora_lily retweeted
@voicearena_ai 's 94% figure is pooled across the evaluation. The best-model result is a different statistic. Keeping them separate matters.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
Your skincare product might look amazing on camera, but can viewers actually hear what you're saying about it?
CapCut PC's Black Friday deal is on its final day: 60% off Pro and 40% off Ultra. For a UGC review with a voiceover, I'd use Edit Pilot to optimize the original audio, then use Enhance Voice if the spoken parts need more clarity. I'd also adjust the voiceover and music separately with Multi-Track Editing.
tinyurl.com/2v23rn22
#capcutpc #capcutmarketing #capcutblackfriday #capcutsale
Aurora_lily retweeted
If Turing TTS encourages better human-baseline tests elsewhere, that would be a contribution even if another method eventually becomes the standard.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
We’ve raised $1.3M from YC & Afore Capital so you never have to type or speak again.
Typing is slow. Talking into your computer is awkward. Subvocal fixes both
Our glasses give you the speed of speech without you making a sound or touching a keyboard.
It tracks your internal speech muscles with non-contact sensing and decodes silent, almost invisible mouthing into text at more than 200 words/min.
Pre-orders are now live at:
subvocal.co/preorder?utm_sou…
The conversation should outlast the Google result. What does the next model teach us about human identification that this one doesn't?
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Aurora_lily retweeted
🚀 GEMINI vs PERPLEXITY: Which AI Comes Out on Top?
Discover 10 powerful prompts to unlock the full potential of both AI tools for research, content creation, productivity, and more!
Test these prompts and find out which AI delivers the best results. 👇
Follow for more AI tips & powerful prompts! 🧵
Aurora_lily retweeted
🚨 BREAKING: Gemini can now handle your entire YouTube research and scriptwriting workflow like a $5,000/month production team — for FREE.
These 8 powerful prompts turn simple channel ideas into engaging, high-retention YouTube scripts in seconds. 🧵
