Add an Audio generator: speech read from the text with the text to speech provider

Problem/Motivation

AI Media makes images, SVGs and, experimentally, videos. Sites also want spoken clips: an intro, a pronunciation, a short announcement. The AI module has a text to speech operation with real providers (OpenAI, Gemini, ElevenLabs), and core ships an Audio media type.

Proposed resolution

• An Audio generator: the text is read aloud by the text to speech provider and saved into a media type with the Audio file source. Off by default, like Video; turned on in the generators table.
• The text field is "Text to speak", with its own help; a generator may now name its field.
• Options: the model, the voice the model lists (the site sets the default and may show the choice), and the format: mp3, wav or aac.
• A sound or video result shows a player where an image shows its preview, and its link reads "Open in a new tab".
• Only an image is asked for alt and title; other media types ask AI for the name alone, one call less per generation.
• Kernel tests, and an e2e scenario on Add media with the Echo provider, which speaks too.

Remaining tasks

• ✅ File an issue
• ✅ Addition/Change/Update/Fix
• ✅ Testing to ensure no regression
• ✅ Automated unit/functional testing coverage
• ✅ Developer Documentation support
• ❌ User Guide Documentation support
• ➖ UX/UI designer responsibilities
• ➖ Accessibility and Readability
• ❌ Reviewed by a human
• ❌ Code review by maintainers
• ❌ Full testing and approval
• ❌ Credit contributors
• ❌ Review with the product owner
• ✅ Update Release Notes
• ✅ Release

User interface changes

Add media > Audio, Generate with AI, after a generation in Gin: the AI name, the text, the player and the facts line.

The Options with Format and Voice shown:

API changes

• The AiMediaGenerator attribute takes optional field_title and field_help.
• AiMediaMetadata::values() takes the targets to fill; the default is all of them.

Data model changes

• ai_media.settings gains generators.audio and editor_options.audio, with a voice option.

Release notes snippet

• Added the Audio generator: speech read from the text with the text to speech provider, in a voice the model lists.

AI-Generated: Yes