OpenAI FM: The Radio of the Future
openai.fm is an official OpenAI demo that demonstrates the text-to-speech capabilities (TTS) of the OpenAI API. It enables generating high-quality AI voices for videos, podcasts, or prototypes. The use of the demo is free, but ongoing production of audio via the API is charged.
Quelle: Custom Presentation
Introduction
openai.fm serves as an interactive website to demonstrate the text-to-speech capabilities of the OpenAI API. Users can enter text, choose a voice, and immediately listen to the result. This is intended as a demo to showcase the quality and simplicity of the integration. The underlying code is on GitHub verfügbar and shows how developers can integrate the Speech API into their own applications.
OpenAI released new audio models in 2025, including improved speech recognition (gpt-4o-transcribe) and a controllable TTS model named gpt-4o-mini-tts. Developers can now specify in detail how the voice should speak. openai.fm serves as an official showcase for these features and is linked on GitHub as an OpenAI repository that uses Next.js code which accesses the Speech API directly.
The demo is freely accessible, but productive use via the API is paid. Prices are charged by OpenAI per token or depending on realtime/audio usage. The current rates are transparently listed in the Pricing List listed.
Functionality & Models
“Text-to-Speech” (TTS) is the process by which written text is converted into spoken language. OpenAI provides several models for this, including gpt-4o-mini-tts. . The available voices carry names like Alloy, Echo, Fable, Onyx, Nova or Shimmer. These voices are accessible via the API and can be integrated into applications.
openai.fm is an official OpenAI demo, which is evidenced by the GitHub Repository, which belongs to the OpenAI organization and describes this purpose. The TTS voices such as Alloy, Echo, Fable, Onyx, Nova and Shimmer are listed and their use is via the API.
Quelle: listoru.hatenablog.com
The intuitive user interface of OpenAI FM enables easy selection of voices and moods for text-to-speech generation.
Costs & Usage
The demo on openai.fm is freely usable and serves to test the API's features. However, for ongoing production use of the text-to-speech functions via the API, charges apply. The API usage is paid, and the prices are in the official OpenAI pricing list listed in detail. Billing is per token or based on realtime/audio usage.
It is misleading to assume that you can permanently produce unlimited free audio with openai.fm. The demo only shows the functions; for production via the API fees apply. Concrete “costs per minute” for each TTS voice vary depending on billing method (tokens, realtime-audio) and model. Flat-rate per-minute prices from forums are estimates and not normative.
Why does OpenAI offer a freely usable demo? Demos lower the hurdle to try new features — for creators, startups and teams. OpenAI shows how voices sound and how easy the integration is. This builds trust in quality and latency before budgets are bound. For quick tests, demos and prototypes, openai.fm is ideal. For productions – such as YouTube voiceovers, e-learning or apps – API costs must be planned.
Quelle: genspark.ai
Professional use of OpenAI FM in various application areas.
Security Aspects & Abuse
There are important security questions around AI voices. OpenAI has not publicly rolled out a powerful voice cloning tool called “Voice Engine” due to abuse concerns, such as identity fraud. Tests run only with partners under strict conditions. This caution also shapes which voices are enabled in the API: preset voices are offered instead of freely clonable identities.
It is false or misleading to claim that OpenAI offers freely available voice cloning for every voice. OpenAI has not generally released voice cloning for security reasons. Media criticism of AI voices often revolves around rights and similarities to real people. A prominent case was the “Sky” voice, paused after criticism by Scarlett Johansson. OpenAI emphasized that the voice came from another actress.
In parallel, third-party websites advertise “free” TTS usage. Pages like openai-fm.com or ttsopenai.com are independent providers with no connection to the official OpenAI domain. They promise free TTS but are not part of the OpenAI platform. For credibility, it is important to check the imprint, privacy policy, limits and the business model of such providers.
Quelle: YouTube
The DevDay breakout shows live how the realtime and audio APIs for voice agents work together — useful context for latency, control, and architecture.
Practical Application
For quick tests, demos and prototypes, openai.fm is ideal. For productions like YouTube voiceovers, e-learning or apps, API costs should be budgeted and rights, disclosure as well as abuse protection considered. Practical tips include using preset voices and avoiding cloning real people. It is advisable to label AI audio in projects and check official documentation and repositories instead of name similarities. Start points are the TTS-Guides and Model Documentation.
Open questions concern the further development of the preset voices and the addition of new languages and accents. It is worthwhile to look at the regularly updated changelogs and documentation. When and under what conditions voice cloning will be broadly released is also an open question. OpenAI points to ongoing safety assessments and partner pilots; a date is not publicly fixed.
Quelle: user-added
Announcement of a new text-to-model model by OpenAI.fm.
openai.fm demonstrates impressively how good today’s AI voices sound – free of charge as a demo, but not as an unlimited production platform. For serious endeavors, the path leads to the official API with clear pricing, preset voices and safety guardrails. Those who work cleanly, link sources, label AI audio and rely on official documentation and repositories can quickly develop a robust workflow from a demo.
Quelle: YouTube
A tutorial explaining the openai.fm interface and typical export pitfalls.