laitimes

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

author:Venture State
The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

Author丨Linfeng

Editor丨Sea waist

Image source丨Picture insect creativity

The number of podcast creators is increasing year by year.

According to Spotify, there will be more than 5 million podcast episodes on its platform alone in 2023, and the total number of active podcasts will range from 3 million to 4 million, covering a variety of genres. Among them, many podcasts are completed by text-to-speech AI, and the voice does not have the artificial mechanical flavor of Siri or Xiao Ai's classmates, and sometimes it can even "confuse the real with the fake".

Now, the Armenian company Podcastle has shined a light on its signature skill, which allows users to clone their own voice and make a voice "skin". Moreover, there are more than 30 AI voice "skins" for users to choose from within the platform, including multiple languages (including dialects).

Founded in 2020, Podcastle has built an AI tool platform for podcast creators, integrating voice cloning tool Revoic, noise reduction tool Magic Dust AI and team collaboration functions, and first opened the market in the United States to complete user accumulation.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

(图源:Podcastle)

Podcastle said in the announcement that it caters to the content creation needs of podcast creators, sole traders, marketers and educators, and that the creator community has grown rapidly, from 150,000 in 2021 to more than 1 million.

The founder, Arto Yeritsyan, is an Armenian entrepreneur who graduated from Yerevan State Engineering University and Stanford Graduate School of Business. He was previously VP of Engineering at PicsArt and Head of Technology at Be2. Currently, its leadership team is dominated by Canva, Prezi, Uber, and Facebook.

In February this year, their latest Series A funding round of $13.5 million was led by Mosaic Ventures, with participation from RTP, P9, Sierra, Andrew Ng's AI Fund, and the CEOs of Squarespace and Moonbug Media, of which Sierra and AI Fund have invested in three consecutive rounds.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

Create your own voice "skin"

Podcast Host surveyed 2,500 podcast creators, and 32% said podcast editing and production was the biggest concern.

Speaking at SpeechTech, Podcastle CEO Arto Yeritsyan said that podcasts have two major technological leaps, one is that text-to-speech technology allows creators to make audio without having to speak, and the other is to create their own clones without having to repeat recordings. Both of these can be achieved with artificial intelligence tools to achieve more efficient creation.

Riding on this technology trend, Podcastle went from being an extension to video podcasting in 2022 and is now a platform for AI suite tools. Users can complete the whole process of recording and audio editing within the platform. The platform focuses on long-form content creation and differentiates itself from its competitors by offering real-time collaboration and artificial intelligence capabilities.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

(图源:Podcastle)

It integrates tools such as voice cloning, noise reduction, text-to-speech TTS, and more, all for a fee. Among them, the noise reduction enhancement feature Magic Dust AI has been recommended many times by review bloggers, which claims to eliminate background noise and equalize audio current with "one click", and convert low-quality recordings to studio level.

Judging from the example audio effect, it removes the wind sound and current sound present in the original equipment recording, leaving the human voice and increasing the sound quality.

Another feature is Revoic, where artificial intelligence generates voices. Users can clone a digital copy of their voice. Users need to record 70 sentences they read and submit voice samples, and within 24 hours after starting Revoice, they can get a voice template that imitates the real voice.

AI analyzes recorded voice samples, learning human intonation, accents, and details to create a digital replica. It's like a voice "skin" that can be cloned and can be used in a variety of scenarios.

According to Arto, Revoic means podcast creators can rely less on in-studio environments and professional recording equipment for high-quality results.

The Verge作者David Pierce使用了The Office的《Dwight's Perfect Crime》的片段文本,并将用他自己的声音在Podcastle克隆了AI语音,比较了几个语音平台的生成效果。

In terms of effect, ElevenLabs' generation effect is the most realistic, with the personal emotions of the voice owner, followed by Podcastle, which is close to the real human voice and slightly weaker emotional expression.

Blogger Feisworld tested Revoic, the human original voice and the AI voice template can hear a clear difference, the AI version of the voice timbre is close to the original, the words are clear, but slightly mechanical, more like reading the original text word by word, the human original voice has an undulating intonation, sonorous and powerful.

"I wouldn't see AI speech as a tool to replace me, but I think it's beneficial for other tasks, such as reading difficult scientific or philosophical texts, and advertising slogans," Fei said. "It would be great to be able to change the tone of voice if it was an AI-generated voice, but at the moment (in 2023) I don't see an option/method to change the tone in Podcastle." ”

Podcastle Voice Clone Revoic does have a suitable scenario for podcast creators who need to record a lot of voice-overs, voiceovers, and special languages/dialects when creating long videos/audios.

In addition to these two features, Podcastle, like most AI platforms, has basic features for processing audio, including speech-to-text, TTS text-to-speech, AI mute removal (automatically removing long pauses), and nonsense detection (automatically removing words like "um", "you know"), etc.

There are 30 TTS template voices (voice skins) to choose from on the platform, and 7,000 copyright-free songs are also included for creators to use. Users can enter text to create a one-person podcast, or they can choose the appropriate skin by paragraph to simulate a multiplayer scene.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

(来源:Podcastle)

The platform enables cloud recording, and if the remote podcast content encounters a network outage, the content can also be saved instantly. At the same time, when it comes to multi-person podcasts, you can use Podcastle's team collaboration to edit in real-time.

Currently, Podcast is not the only one that is using AI to empower podcasts. Riverside focuses on fast audio editing, Descript-integrated SquadCast improves quality and efficiency, and Listener.Fm uses artificial intelligence for program annotations, titles, and descriptions. Reddit users commented that Podcastle has the simplest interface, a lower platform and error rate, and a more friendly subscription price of $11.99/month (Descript $12/month), which has attracted many creators to try it.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

(图源:Podcastle)

By the end of 2023, Podcastle said it had more than 12 million podcast episodes and content within the platform.

Grew to 1 million users in 2 years

Podcastle's good product operation and user growth are related to founder and CEO Arto Yeritsyan. Former colleague Tammy commented: "He has what it takes to be a successful entrepreneur, technology, product knowledge, people and business acumen. ”

He's got a lot of skill. Arto listed 50 skills on LinkedIn, covering software, programming, executive management, and more, and was named one of the 30under30 Armenian tech talents by Hive Ventures in 2020.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

(图源:Arto Yeritsyan领英)

Born in Yerevan, the capital and economic center of Armenia, Arto graduated from the Armenian National Engineering University and spent one year at Stanford University's Graduate School of Business. After graduating, Arto worked as an engineer at Be2, one of Armenia's leading technology companies, and became a technical director in two years.

At that time, as a part-time worker, Arto firmly believed, "If I perform my best, I will be recognized." In their letters of recommendation, former colleagues labeled Arto as "clear-minded" and "cognitively clear." Later, he seized the opportunity to work for PicsArt, Armenia's largest unicorn, for seven or eight years, rising from director of product development to vice president of engineering, and becoming a key figure in managing a 300-person engineering team. Arto also worked as a consultant for PicsArt for 2 years after leaving.

Around 2020, many entrepreneurs became interested in AI, and Arto was no different. He was dyslexic when he was in school, and much of his knowledge had to be absorbed by hearing. In his opinion, audio is the simplest and most direct way to communicate.

The podcast is an in-depth interview presented in the form of audio, which is different from the lecture of one-way knowledge indoctrination, which outputs in-depth views and unique dismantling through chat interaction.

So, while working at PicsArt, he came up with the idea of making a podcast tool. He is a company executive, and he is busy like a spinning top every day, and he has imagined a simple plug-in to digest articles and blogs for him in the form of podcasts. In June of the same year, he and his friends founded Podcastle to explore entrepreneurship.

Prior to PicsArt, he also ran a company as a co-founder of Coding Records. From a part-time worker to a manager, Arto has gradually accumulated experience in identifying and employing people throughout his career. As a result, he recruited 3 like-minded founding members for what would become Podcastle. Aram, former chief software development engineer at Policis, Arsen from museum marketer to WIC director, and Vardan a veteran of Webb Fontaine products.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

(图源:Arto Yeritsyan领英,依次为Arsen、Arsen、Vardan、Aram)

However, none of the four of them had ever dive into podcasts before, so they stepped on a pit in the first place. At first, Podcastle focused on text-to-audio features, such as Arto's original vision of a Chrome extension that would turn any article into a podcast in seconds, gaining its first users, but it was difficult to take the next step to attract professional podcasters.

"We only took into account the consumption of audio content, not the needs of podcast creators. "In 2022, after Arto decided to shut down the plugin and make up for the podcast content creation and editing posts, he invested all his resources in developing digital creation and audio enhancement, and developed editing and transcription suite tools for creators. That's when their goal really came to mind: to create an all-in-one platform that would make it easy for beginners and professionals to get high-quality audio with the help of AI.

The AI podcast tool has more than one million users in 2 years, and Ng Enda has invested in 3 rounds

(图源:Podcastle)

They tweaked the product design and developed audio editing features such as multi-track recording, auto-equalization, and dynamic fade-in/fade-out of sounds. Users can record individual or group podcasts with up to 10 people within the platform, and a real-time podcast collaboration feature was launched in September 2023.

In addition to satisfying the needs of professionals for sound quality, Arto's target audience also includes inexperienced amateur podcasters. "Everybody has a story in their mind, but they don't have the specialized equipment, they don't know how to read it, and how to retain an audience. But anyone should have the confidence and ability to make their voice heard. ”

To solve this problem, Arto automates the whole chain of podcasts from consumption to creation, and even takes care of the voice. He revealed that providing users with a rich "sound skin" is one of Podcaste's strategies to achieve profitability. In addition to providing more than 30 fixed voices, Podcastle's text-to-voice system has also developed Voice and additional voices to provide paid subscriptions. In addition, Podcastle's official blog is a collection of podcast experience posts, recommended tool posts, and case studies.

In order to truly achieve a "one-stop service", Podcastle's hosting platform provides creator users with podcast RSS feeds to ensure that users maintain listener stickiness. For subscribers, they can record 20 hours of 4K video podcasts per month, and the platform can do a simple background blur.

Podcastle accumulated about 200,000 users at the end of 2021, and after product adjustments in 2022, it ushered in rapid user growth, and by the end of 2023, it had more than 1 million users in its community.

Financing in the context of the war began

Armenia has a poor streaming environment, and people rarely talk about social issues or make public comments in English.

Arto chose to launch Podcastle in the United States. He found that young people in the U.S. don't access information from traditional media, but from podcasts or other sources they trust, making it a better environment for podcast creation: "They appreciate and understand structured discussions, and they don't stay in one camp." ”

In order to attract users, Arto has maintained a free plan for text-to-speech and language transcription for the first 2 years in Podcastle. The long-term free program needed funding, and Arto had to raise money.

He and his friends are all set, and Arto has built up a multi-national industry network through his work. However, at that time, Armenia was torn between war and peace, and the outbreak of the pandemic led to a widespread economic depression. This made it difficult for Armenian companies to secure funding, and Arto spoke to 50 investors several times, only 2-3 of which worked.

In an interview with Rearrange, Arto revealed that he saw the opposite business ecology in Armenia in the increasingly involuted global wave of entrepreneurship. "Startup executives or founders are trying to help other people by maximizing their impact as much as possible. ”

PicsArt, a strong unicorn in Armenia, has given other tech startups in the country a lot of fundraising. First, Armenia has a lot of outstanding scientific and technological talents, and PicsArt has made it possible for them to enter the international scientific and technological arena. The second is that PicsArt is committed to making products that users all over the world love, and is working hard to get funding.

"In our country, most companies are helping each other out and are excited about the success of other companies," Arto said. It's rare that they (PicsArt, Krisp, etc.) will let a lot of investors from Silicon Valley know about Armenia. He lamented that although Armenia is small, it can play a role in the larger world through cooperation.

At the end of 2020, Podcastle received a $1.75 million round led by US-based VC Sierra Ventures after achieving its first organic user growth.

Podcastle is the second podcast company invested by Sierra Ventures, the other being Himalaya FM, which was valued at more than $3.5 billion at the time. They came across Podcastle in Armenia when they were investing in Krisp, "Krisp grew from $0 to $4 million ARR in 1 year, and given that Podcastle is also Armenian, I believe they can build a capital-efficient business with talent and projects." ”

It is worth mentioning that as early as this round of financing, Podcastle received an investment from Andrew Ng's AI Fund.

Arto never shied away from talking to investors about his country being at war, which made most of the investors who were interested in his project immediately shut down. But the AI Fund wasn't intimidated, "they were interested and believed we could grow on our own." ”

Simon Levene, co-founder of Mosaic, the latest round of lead investors, is also bullish on them, "Arto's products are showing a trend of organic growth, and this growth will accelerate in the coming years. ”

The number of people listening to podcasts is also increasing year by year. According to The Infinite Dial, more than 6 in 10 Americans (over the age of 12) are podcast listeners, a figure that jumped to 73% at the start of 2022. Demand Sage also has data that the global podcast audience will reach 504.9 million in 2024.

At present, in Arto's view, there are two biggest trends in the podcast industry, the biggest direction is sound quality improvement, using AI tools to generate "pseudo-high-quality audio" for any audio, and the other general direction is marketing, high-fidelity AI voice can assist in the distribution of a large number of advertising slices, helping enterprises to share and drain.

Podcastle's announcement shows that in addition to accelerating the development of AI tools, it will also need to expand its product range after the investment. To this end, Arto has recruited a lot of talent. Some time ago, Allan, the former vice president of Canva, became the chief commercial officer, and the leadership team was joined by veteran employees from Prezi, Uber, and Facebook.

Read on