Each demo runs on the same API and compute you build with. Open one, then read the docs.
Built by the GMI Cloud community and GMI DevRel. 49 demos, each one a small app that calls GMI inference and returns a result.
One sentence in, one image out: Gemini 4 Argon writes the image prompt, Nano Banana 2.1 renders it on GMI Cloud. Covers, wallpapers, ads, photo edits.
Spotify history → Wrapped-style slideshow: Claude Opus 4.7 finds your eras and writes the roast, Gemini 3 Pro Image paints each chapter.
Paste a topic or article: MiniMax M3 writes a two-host episode, Speech 2.8 voices it and Music 3.0 scores it with an original theme. By zkasuran.
Theme in, songs out: MiniMax-M2.7 drafts title, lyrics and style per track, Minimax Music 3.0 renders them. Local library, 3D listening room, by snekkenull
Drop objects into a photo: Gemini 3.1 Pro reads the scene's light and perspective, Gemini 3.1 Flash Image composites them in as one shot.
Audio description for blind viewers: MiniMax M3.1-Flash reads the frames, writes one line per silent gap, the browser speaks it in the gap. WebVTT out.
Turns a GitHub repo into a two-host podcast: Qwen3.8 writes the script, two GMI TTS voices read it, the browser mixes the audio.
Drop in an ad, pick one of six Asian markets, and get a culture-aware strategy, native-language copy and a localized ad image. by Hyeseong-Ju
NVIDIA's CFO commentary plays as a ticker while Step 5 Preview ties every number to the 10-Q: all 43 filing pages as images plus text in one request.