OpenAI image-2.5 hands-on: sub-$0.01, 23s generations clearly beat Muse
tobowers · x · 2026-09-09
Developer tobowers tested OpenAI's new image-2.5 model and calls it a major upgrade: low-quality generation takes 23 seconds at under $0.01 per image. Compared against Muse at the same price but 40 seconds generation time, image-2.5 produces noticeably richer images with better world coherence.
More from Multimodal
- User Claims 'GPT 6 Astra' Turned a Cartoon Clip into a Full Blender 3D Scene — petewoodbridge · 2026-09-09
- GPT-6 Astra transcribes piano videos into synced sheet music instantly — TianbaoX · 2026-09-09
- 113 TTS eval papers mapped: WER rewards kill prosody, UTMOS doesn't travel, and MOS scores aren't comparable — Slight_Republic_4242 · 2026-09-09
- Saudi dev crafts ChatGPT prompt for low-credit motion graphics, doable from a phone — aziz4ai · 2026-09-09
- Full text-to-video prompt template for raw found-footage alien planet vlogs — techhalla · 2026-09-09
- VoiceStudio open-sources a local ElevenLabs-style stack: voice clone, dubbing, 646 languages — thisguyknowsai · 2026-09-09