Back to changelog
Product
Flare defaults to OpenAI for text, images and speech
OpenAI adapters, explicit local live mode, updated credit prices and a paid integration smoke check.
- Chat, document summaries and prompt enhancement default to
gpt-6-luna; images use GPT Image 2.5 Flare, and speech usesgpt-4o-mini-ttswith Coral by default. - Configure server-side
OPENAI_API_KEY. Local development still defaults to mock mode; real OpenAI calls useAI_MODE=live. Only optional local Workers AI calls requireWORKERS_AI_REMOTE=true. - Default images cost 20 credits each and output PNG. Speech costs 5 credits per 1,000 characters, rounded up with a minimum of 1, and outputs WAV. Chat remains 1 credit and summaries 5.
- Video keeps its existing provider. The OpenAI Videos API has closed; the Speech API adapter needs migration before January 6, 2027. See the lifecycle notes.
- The source includes an opt-in paid OpenAI smoke script. Provider calls and the local authenticated Worker flow were verified, including private assets, audio playback and credit settlement. This is local integration evidence, not a production deployment report.
The landing page's existing media are historical Flare samples; the displayed model settings describe the current defaults.