Pocket FM, an Indian-based audio series platform manufacturer, is aiming to become Netflix audio. In other words, the company intends to match the audio series and hundreds of episodes to users' preferences. To do this, you need to release your content quickly. This is what we are now looking at AI.
The Lightspeed-Back startup offers writers an AI toolset that can help them to suggest a better ending of an episode or do things that make the story more appealing. The hope is that the tool speeds up the story writing process.
Pocket FM is already using AI tools such as elevenLabs to generate voices for the audio series. We also internally tested AI tools for writing and adaptation support.
Pocket FM founder Rohan Nayak said it would take time to finish the episode as he deploys AI tools to all writers.
Image credit: Pocket FM
Use a writing tool called Copilot to allow writers to create stories.
Copilot can convert narrative-based lighting to dialog-based lighting for a specific segment. You can also “beat analysis” to shape the sentences to make audio series of a particular genre more attractive. The tool has basic chatbot style writing features, including “abbreviation”, “expanding”, and the ability to generate text via prompts.
To build Copilot, the company looked at thousands of hours of data points to understand why users are more involved in a particular storyline in a particular genre.
Based on that, we added a writing suggestion feature designed to increase conflicts between characters and recommend endings to make the episode more exciting. AI can also suggest background effects tags that can be used while creating audio.
This tool automatically generates the BIOS for the characters, their relationships, summarizing plot points for various episodes, allowing creators to see these details during creation.
Copilot also has a review tool that checks plot points, grammar and leaves qualitative feedback through comments about episodes.
Under the hood, Pocket FM is training a smaller model to maintain the context of the character arc and relationship story, along with story consistency. Additionally, using user signals, startups tweak their AI to add drama to the story.
International expansion and localization plans
In addition to the arrival of AI tools, Pocket FM has launched adaptation tools for various markets that not only translate text from one language to another, but also change names and phrases that are more appropriate for the culture of the region.
The company first debuted the tool as part of a German Copilot Suite earlier this year to transform stories from other regions after reportedly struggling to attract users in European countries last year.
Nayak said the company saw the significant results of the trial and saw its monthly in-app revenue constantly rise, exceeding $700,000 in June.
Image credit: Pocket FM
“When we started expanding into a new region, it took us 12-18 months to be meaningful in that market. We need at least 1,000 hours of content to start getting users and scaling the market. This can be done in less than three months,” says Nayak.
This tool has increased the productivity of writers in the German market by up to 50% in terms of show output. Additionally, the tool helped the company create error-free drafts of the show, resulting in higher user retention rates for the audio series.
In the US, a series created with the help of these new AI tools is currently contributing to 10% of Playtime. Additionally, these shows have generated $7 million in revenue over the past 12 months, reducing production costs by 2-3 times.
Build technology to expand content generation
With the adoption of different AI capabilities internally, Pocket FM was able to quickly scale content. The startup said it will launch nearly 1,000 pilots per month. And only a huge amount of content will make some of them a hit.
But audio shows are just a small part of it. The company is already working on tools to convert stories into comic strips on the Pocket Toons platform. Additionally, Nayak said the video is also a format that the company can explore. The startup, which raised more than $199 million in funding across the round, is also experimenting with microdrama apps.
Image credit: TechCrunch (screenshot)
By next year, Pocket FM hopes to release its own singular language model (LLM). It incorporates a variety of tools, such as writing support, adoption, dramatization, and story context retention, based on data collected from the show. The company said that switching to its own LLM would not require training a large number of small models for individual features.
Potential drawbacks of AI
There were side effects of AI adoption.
Pocket FM has already fired people who are employees or contractors in multiple rounds over the past 12 months. There have also been reports that writers are seeing reduced returns over time. And the company faces lawsuits in California over employment and wage issues.
“Like most content-driven industries, we work with a diverse network of writers, voice artists and production partners on a project-based basis to coordinate resources for each market. AI has had a minimal impact on our core creative community. Instead, it paved a new pathway to expanding reach and output.
There are also questions about quality. The company measures quality by the number of shows held.
The basic argument is that new AI tools can also serve as writers' rooms for solo creators, allowing them to create more content at a rapid speed. Furthermore, based on numbers, writers can quickly edit stories with the help of AI. However, these tools can very well direct “AI Slops,” a content generated by low-quality AI, into the platform, which can affect user recommendations, making it difficult to discover good stories.
Pocket FM claims that despite AI helping them, stories with solid structures gain popularity.
The company noted that all content is reviewed by an AI-powered moderation framework to ensure quality and originality. It also insists on AI moderation checks such as duplication, copyright issues, content health and other quality measurements before approving the audio. Each show receives an equal push, and user engagement ultimately determines the ranking of the show.
Another concern is that writers can overrely rely on AI over time.
In Germany, AI writes more content than humans per show with a selected title. Pocket FM's plan to deploy more AI tools could increase the amount of AI writing content. And that could also raise expectations of burning more shows. Unless user recruitment also rises rapidly, average returns can drop.
The company did not directly address TechCrunch's questions about returns, but said that AI tools would help speed up the author's work and edit episodes based on numbers and audience feedback. This means you can make targeted improvements instead of doing a complete rewrite.
“This faster content creation does not necessarily dilute quality or relevance. The author's role simply shifts towards more productive editing, refinement and maneuvering the output,” the spokesperson said in a statement.
We are constantly aiming to evolve and you can help us by providing insights into TechCrunch and your perspective and feedback on our coverage and events! Fill in this research to let us know how we are doing and get the opportunity to win an award in return!