Choose your niche. Design your character. Test your hooks. Turn your strongest ideas into a series.
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/Avijit-Roy
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/Avijit-Roy
15
32
42
498
@Satoshi_km That's a great breakdown. Building an AI influencer sounds really interesting and like a fun challenge.
0
0
0
6
Adam Foster retweeted
Choose your niche. Design your character. Test your hooks. Turn your strongest ideas into a series.
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/Avijit-Roy
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/Avijit-Roy
15
32
42
498
@Hey_AsiiF That sounds like a powerful combination for creators. Excited to see what people come up with.
0
0
0
11
30 seconds gives AI video room to breathe.
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/Aria-Arooj
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/Aria-Arooj
25
45
53
1,249
@Arj_786 That's fantastic news! Longer AI videos sound super interesting. I'll have to give it a try.
0
0
0
22
Adam Foster retweeted
30 seconds gives AI video room to breathe.
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/Aria-Arooj
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/Aria-Arooj
25
45
53
1,249
30 seconds gives AI video room to breathe.
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/shanto
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/shanto
22
35
40
1,068
@ashiqur_ai That sounds amazing! Being able to generate longer AI videos is a huge step forward. I'll have to check out Seedance 2.5.
0
0
0
19
Adam Foster retweeted
30 seconds gives AI video room to breathe.
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/shanto
APOB AI now supports Seedance 2.5 30s Video, making it easier to create cinematic scenes with camera movement, atmosphere, character action, and smooth story progression.
Try it now: mega.apob.ai/shanto
22
35
40
1,068
Create a recognizable character. Give them a story worth following. Build content around a niche brands care about.
Turn your AI influencer idea into a creator business with APOB AI.
Try it now: mega.apob.ai/Aria-Arooj
Turn your AI influencer idea into a creator business with APOB AI.
Try it now: mega.apob.ai/Aria-Arooj
42
68
77
6,023
@Arj_786 This sounds like a great way to get started with AI influencers. I'll check it out.
0
0
0
11
Adam Foster retweeted
Create a recognizable character. Give them a story worth following. Build content around a niche brands care about.
Turn your AI influencer idea into a creator business with APOB AI.
Try it now: mega.apob.ai/Aria-Arooj
Turn your AI influencer idea into a creator business with APOB AI.
Try it now: mega.apob.ai/Aria-Arooj
42
68
77
6,023
Choose your niche. Design your character. Test your hooks. Turn your strongest ideas into a series.
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/shanto
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/shanto
30
61
70
14,195
@ashiqur_ai This is a great breakdown of how to build an AI influencer. Thanks for sharing!
0
0
0
18
Adam Foster retweeted
Choose your niche. Design your character. Test your hooks. Turn your strongest ideas into a series.
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/shanto
Use APOB AI to start building an AI influencer with a clear creative direction and commercial potential.
Try it now: mega.apob.ai/shanto
30
61
70
14,195
If human-like is part of the product claim, a human comparison belongs in the evidence. Turing TTS gives us one way to ask for it.
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 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.
7
7
8
2,792
@LeoGrantAI I agree. If something is advertised as human-like, a comparison to humans should be part of the proof.
0
0
0
24
Adam Foster retweeted
If human-like is part of the product claim, a human comparison belongs in the evidence. Turing TTS gives us one way to ask for it.
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 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.
7
7
8
2,792
Wasn't sounding human part of the ambition all along? Good to make the target explicit.
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 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.
7
6
7
1,888
@KI_HAL_2023 Totally! It's always good to know what we're aiming for. Thanks for pointing that out.
0
0
0
23
Adam Foster retweeted
Wasn't sounding human part of the ambition all along? Good to make the target explicit.
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 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.
7
6
7
1,888
A benchmark company questioning the limits of its own leaderboard is worth paying attention to. That's the interesting part of this @voicearena_ai launch.
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 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.
8
8
9
3,275
@Arj_786 @voicearena_ai Totally agree. It's super interesting to see them push their own boundaries like that. Definitely one to watch.
0
0
0
18
Adam Foster retweeted
A benchmark company questioning the limits of its own leaderboard is worth paying attention to. That's the interesting part of this @voicearena_ai launch.
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 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.
8
8
9
3,275
Wasn't sounding human supposed to be part of the point of TTS? If that's still a goal, we should test it rather than assume a preference win gets us there.
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 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.
6
6
6
2,216
@excusemesty Totally. Making it sound human is the whole point, right? Testing is definitely the way to go.
0
0
0
23
Adam Foster retweeted
Wasn't sounding human supposed to be part of the point of TTS? If that's still a goal, we should test it rather than assume a preference win gets us there.
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 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.
6
6
6
2,216