Coqui, a conversational AI startup, on Wednesday (January 3, 2023), announced that it is shutting down its operation, most likely due to funding crunch and monetisation challenges. 

Founded in 2021 by Eren Gölge, Josh Meyer, Kelly Davis, Reuben Morais, the company specialises in building open source models and applications in the area of quick voice cloning, text-to-voice, etc. The former employees of Mozilla, left the company after it stopped developing their own Speech-to-text engine, DeepSpeech to begin Coqui.

The shutting down of the company comes as a surprise as it had raised a funding of $3.3M in March last year. Morais announced the end of the company to their subscribers through e-mail in early December saying, “We have made the difficult decision to discontinue our paid SaaS web application and REST API services. This means that our servers will be going offline in 14 days, on December 11th, 2023.” 

The company had also announced the launch of XTTS on September 14th last year, open sourcing the Text-to-speech (TTS) foundation model. It is capable of generating speech in 13 languages and features voice cloning, including cross-language voice cloning. Coqui, in collaboration with Heather Meeker, introduced the Coqui Public Model License (CPML) with this release, marking a new approach in model licensing. 

According to Gölge, its open source foundation models would still be available on GitHub and HuggingFace for developers to use.

Tough year for start-ups?

The founders gave no specific reason as to why it decided to shut down the company but Gölge responded to a user on GitHub, saying they wanted to focus on “open-source and model development.” 

Coqui operated primarily as an open-source project, focusing on developing and providing tools and resources for automatic speech recognition (ASR) and Text-to-Speech (TTS) technologies. 

The company democratised access to speech technology by offering free and open-source ASR models, datasets, and training pipelines. This open-source model allowed for wide accessibility and encouraged innovation, as users could modify and distribute the technology freely.

Eventually, Coqui provided commercial services alongside their open-source offerings. These services include consulting, custom model development, and training services to assist businesses in integrating speech technology into their products or operations. This dual approach of open-source and commercial services allowed Coqui to cater to a broad range of users – from individual developers and researchers to large corporations seeking custom solutions.

Its fair pricing which made them popular among users, could be one of the reasons why it was hard to sustain as a business. Christopher Manning, the Director of Stanford tweeted in response to the news, “The lesson is probably that it’s just really hard to succeed as a standalone TTS (or similar lang tech) company.”

It has been a difficult year for AI startups in the past year. “The most difficult year for startups in at least a decade,” Peter Walker, head of insights at Carta, a company that provides financial services for many Silicon Valley, wrote on LinkedIn.

The venture capital industry is currently facing a challenging time. At the forefront of these challenges are the notable shifts in leadership at prominent venture capital firms, coupled with a slowdown in the pace of deal-making. 

In response to this, seed stage startups receive smaller funds, while the venture capital sector is taking a more cautious approach adopted by limited partners in their investment strategies. 

There is a lot of pressure on new startups to quickly build a business model and be profitable. Sridhar Ramaswamy, Senior VP of AI and the former founder of Neeva told AIM, “Some companies are raising a lot of money, but I think they have to quickly show the business value of what they are doing.” His own company, Neeva, which began as an ambitious endeavour in the search engine space was quickly acquired by Snowflake after it got shut down abruptly.