MAI Playground (https://playground.microsoft.ai/) provides access to Microsoft’s AI models: image generation, audio transcription, text-to-speech, and multimodal conversation.

4 models available:

• MAI-Image-2.5 generates and modifies images from a textual instruction or a source image (did not work during testing). A Flash version is available.
• MAI-Transcribe-1.5 converts an imported or recorded audio file to text, 25 languages.
• MAI-Voice-2 synthesizes text to audio with 45 voices, 19 languages, and 8 styles.
• DuoAI is an experiment combining transcription and text-to-speech: the user interacts in real time with two voice characters, Luna and Leo.
  Two modes: free conversation and interactive step-by-step story building. Works only in English but understands French.

A reasoning language model, MAI-Thinking-1, has been announced and will be the subject of a new paper.

Free access and seems unlimited.