The article AI generated podcasts: If talks have never taken place from Valerie Wagner first appeared on Basic Thinking. You always stay up to date with our newsletter.
Various tools use artificial intelligence (AI) to convert text into sound or sound into text and create synthetic voices. This allows podcast talks that have never taken place. But AI generated podcasts also reveal risks in addition to opportunities.
In November 2023, Seven.One Audio published the Podcast series “Gebrüder Glittch”, a fully AI-generated production. The aim of the experiment was to use the potential of AI in marketing. While weather and traffic reports are predicted in the ARD Audiolab Ki, “The Rock- Radio Helgoland” runs completely Ki-controlled.
AI in the media production-generated podcast: If talks have never taken place
An example of AI-based creativity is that Podcast Sheldon County By James Ryan, doctoral student at the University of California in Santa Cruz, who went online in 2018. Here AI takes on all the tasks: she writes stories, taught it to the realistic sounding computer voice “Justine” (Amazon AWS) and offers users: Inner individualized, endless podcasts. The simulation depicts an American small town with complex characters and social structures that can always be combined again. Ryan sees a foretaste of a future in which computers create novels and TV series.
Seven.one audio experimented with Gebrüder Glittch (2023) a AI generated fairy tale podcast. The AI takes over storytelling, language synthesis and even cover design. Nevertheless, the company emphasizes that AI cannot replace human creative sparks. Especially in the event of voices there is often a feeling of strangeness. Ki is particularly useful in brainstorming and marketing: Seven so individualized. One with AI advertising campaigns, for example for automobile manufacturers, who create 300 different spots for local markets.
In Switzerland the Digital agency Netgen with Ai minutes developed an almost completely AI-generated podcast. The team around Dennis Oswald and Amar Delić uses numerous AI tools: from the conception to text generation to the synthetic moderator voice “Lily”. A tailor-made GPT assistant, Midjourney for visual elements and python for automation complete the project. The weekly podcast has been looking back on AI developments since December 2023 and stimulates the discussion about social and moral borders.
A pioneering project is The Rock – Radio Helgoland,, The Thore Laufenberg operates. After the departure of the employees, he developed a AI system that researched the topics, writes texts and presented them with cloned voices of former moderators. Even the voice of a deceased colleague can be heard – with the consent of the relatives. The Helgolands react mixed: While some welcome the concept, others criticize the control from Bremerhaven. Laufenberg sees opportunities in AI, but also risks, for example for jobs. However, the technology cannot yet generate emotions or original content.
AI changes podcast production
Transcriptions that used to take hours do tools such as Whisper or Aiko faster today – precisely and with correct punctuation. Artificial intelligence also supports with research, finding them and scripting. It formulates interview questions, creates discussion threads and proposes episode structures.
In audio production, tools such as Auphonic with AI optimize the sound quality, remove noise, adapt volume and eliminate breaks or filler words. It also creates translations, subtitles, audiograms and marketing material. So lay people can also produce podcasts. For production companies and marketing teams, AI lowers the costs, accelerates processes and often makes recording studios superfluous.
Current AI tools for audiogenization
Tools like Speechify,, Play.ht,, Descript and Notebootklm Support podcast production in different ways. While Speechify, Play.ht and Descript specialize in language synthesis, NudenbooKlm focuses on understanding of text and knowledge management. All four services are cloud-based and offer subscription models.
Speechify, Play. ht and descript:
- AI-based text-to-language technology
- Libraries with different voices
- Adaptation options for votes
Differences:
- Speechify: Developed for people with reading difficulties, offers OCR technology and is available as an app, browser expansion and desktop application.
- HT: Specialized in voiceovers for marketing and the media, enables vocal clones.
- Descript: Comprehensive audio and video editing software with functions such as overdub and text-based audio processing.
NotebooKlm, a AI tool from Google, allows users: to upload and make content such as text files, PDFs or audio files. It summarizes complex materials, answers questions with source information and transforms raw data into structured formats such as study leaders, chronological overviews or audio summary. The target group are researchers, students and knowledge workers.
Text-to-speech and voice cloning
Text-to-speech converts texts into spoken language. A distinction is made between language reproduction, which is based on recorded recordings, and language synthesis, creates the language purely mathematically. Modern systems combine natural language processing (NLP) for text analysis with digital signal processing (DSP) for voice output.
This naturally creates sounding voices. Voice cloning copies voices digitally and reproduces them deceptively real – even from short voice recordings. This technology is used in the film industry and increasingly in podcasting.
Opportunities for AI generated podcasts
Artificial intelligence also supports lay people in podcast production. Tools transcribe audio files to text in no time and make podcasts more accessible – for humans, but also for search engines. Important: The transcript should be converted into a structured article, nobody reads a lead desert, and no search engine.
AI helps with research, finds topics for new content, structured, formulated questions, improves audio quality and provides marketing material. All of this at the push of a button and from existing content and documents. This reduces AI costs and accelerates processes.
With the help of AI, fictional interlocutors can be created that shine with specialist knowledge. Or historical personalities can be “brought to life”. NotebooKlm, for example, only delivers the audio files in English by default, with one promptly it can also translate the sound file in German or any other language.
This makes a podcast more accessible and other target groups can be achieved. Artificial intelligence does not need a break, no vacation, no food, it can create audio content 24/7.
Risks of AI generated podcasts
But NotebooKlm also creates conversations that have never taken place. I recently conducted an interview for a report in the local newspaper. The tool created a 13-minute podcast from it that I never recorded.
At the beginning I told about pictures of women in clothes from cats, which are also illustrations that do not exist. They are invented because AI hallucinated.
Many podcast producers speak that you can get into the ear of your listener with a podcast. And so it is. Good sound is essential for good podcasts. If the tone is wrong, it hisses and crunches, the interlocutors sound as if they had spoken in a sheet metal box, listeners break off. The same applies to monotonous narrators. Artificial voices have no ups and downs, no stalls, no breathing, no breaks. Nothing human.
I couldn’t completely listen to the “brothers Glittch” podcast, the artificial voice bothered me. Everyone who has spoken to Alexa, Siri or Cortana for a long time knows that. It just sounds artificially, the one -dimensional voice, the answer “I didn’t understand that”. Only in the March announced Amazonto save the voice commands in the cloud in the future, instead of so far.
The reason: “Since we expand Alexa’s skills with generative AI functions based on the computing performance of the safe cloud of Amazon, we have decided not to support this function.” What is not there: and to probably train the AI with the customer voices. Privacy and right to your own word adé!
We People recognize us on sound ours Voices,, and our Conversations have a human clay – also in the Internet. Artificial intelligence offers many opportunities, especially in audio processing. But it also harbors risks such as the voice cloning. These technologies can lead to abuse, spam and phishing and bring more damage than benefits. There is nothing to be said against using artificial intelligence, but it must be marked and benevolently used.
Also interesting:
- Artificial intelligence: 7 AI podcasts that you should know
- Hard: New AI creates realistic pictures – faster than ever before
- Ki-token: The smallest linguistic unity of artificial intelligence
- Artificial photosynthesis: New process gains energy from waste
The article AI generated podcasts: If talks have never taken place from Valerie Wagner first appeared on Basic Thinking. Follow us too Google News and Flipboard.
As a Tech Industry expert, I believe that AI-generated podcasts have the potential to revolutionize the way we consume audio content. With advancements in natural language processing and machine learning, AI algorithms can now create highly engaging and informative podcasts that are tailored to the interests and preferences of individual listeners.
While traditional podcasts are typically created by human hosts and producers, AI-generated podcasts have the advantage of being able to quickly generate and distribute content on a massive scale. This can help to democratize podcasting by giving a voice to individuals and communities that may not have had the resources or platform to create their own content.
However, there are also some potential drawbacks to AI-generated podcasts. One concern is the potential for algorithmic bias, where the AI may inadvertently perpetuate stereotypes or misinformation. Additionally, there may be ethical considerations around the use of AI to create content that mimics human speech and emotions.
Overall, I believe that AI-generated podcasts have the potential to be a valuable tool for expanding the reach and diversity of audio content. However, it will be important for creators, platforms, and regulators to carefully consider the ethical implications and ensure that AI-generated podcasts are used responsibly and transparently.
Credits