> ## Content Index
> Fetch the complete content index at: https://nordisk-ai.ghost.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# OpenAI Launches Realtime API and New Tools for AI App Developers
- URL: https://nordisk-ai.ghost.io/openai-launches-realtime-api-and-new-tools-for-ai-app-developers/
- Published: 2024-10-07T10:57:57.000Z
- Updated: 2026-08-27T17:21:19.000Z
- Description: OpenAI unveiled its Realtime API and other tools at DevDay 2024, giving AI developers new ways to build interactive voice features. While OpenAI continues to introduce innovative tools, some of the most anticipated products were missing from this year's event.
- Author: Anders Hagen
- Tags: AI Services and Tools, AI Development, API, Devday, DevDay 2024, OpenAI, Realtime API, #Migrated-1787841634512, #wp, #wp-post, #Import 2026-08-27 14:41, #Import 2026-09-07 13:41

Despite recent weeks' events - including several executive departures and new funding rounds - OpenAI has recently unveiled a range of new tools for developers at its DevDay 2024\. Among the most notable announcements is the "Realtime API," now available in open beta, enabling fast AI-generated voice responses in apps.

### Realtime API Paves the Way for AI-Powered Voice Experiences

The new Realtime API lets developers build apps with AI-powered voice features that respond with low latency. With six unique voices, different from those used in ChatGPT, developers can build interactive voice experiences. External voices, however, can't be used for copyright reasons. OpenAI demonstrated this technology with a travel-planning app where users could talk directly to an AI assistant that gave quick answers and even placed restaurants on a map in real time.

### AI and Ethical Questions Around AI-Generated Voices

With the ability to simulate phone conversations without requiring the AI voices to identify themselves, OpenAI has faced criticism. Instead, it becomes the developers' responsibility to include such disclosures, something that may be required under new legislation in California. Developers can use services like Twilio to integrate calling features, even though OpenAI's technology itself can't directly place calls to businesses.

### New in the API: Vision Fine-Tuning and Prompt Caching

Another new feature is vision fine-tuning, which makes it possible to use images together with text to fine-tune GPT-4o. This can improve the model's performance on tasks where visual understanding is critical. OpenAI does, however, have clear safety rules around which images are allowed, with restrictions on, for example, violent or copyrighted images.

OpenAI also introduced prompt caching, letting developers store frequently used context between API calls, which can cut costs by up to 50%. The feature is similar to one Anthropic recently launched, offering savings of as much as 90%.

### Model Distillation Enables Smaller, Cheaper AI Models

To keep up with the competition, OpenAI is now offering model distillation, where larger AI models are used to fine-tune smaller ones. This can deliver cost savings by letting developers run simpler AI models like GPT-4o mini while retaining the performance of larger models. To help developers analyze these fine-tunings, OpenAI has also launched a beta tool for performance measurement.

### No New AI Models at This Year's DevDay

Despite several new tools and features, this year's DevDay was missing some of the most anticipated releases, such as the full version of OpenAI o1 and the new video model Sora. OpenAI has also held back information about the GPT Store, leaving developers with questions about what's really happening behind the scenes.