Fix the seed for reproducible text-to-image: free course demos NVIDIA's open model Sana
Al_Grigor · x · 2026-11-17
Without fixing the seed, the same prompt yields a different image every time. A new lesson from the free Production AI Agents for Generative Media course shows instructor Mikhail Rozhkov running Sana, NVIDIA's open text-to-image model, on a cloud GPU to generate a greeting card image from a notebook.
- The lesson "First Media Generation in a Few Minutes" demonstrates fixing the seed so every run returns exactly the same picture, in about a second each
- The course started this week and is still open to join
More from Multimodal
- AI-generated cyberpunk loner cityscapes show off cinematic composition skills — Kyrannio · 2026-11-17
- Chalkboard scientific diagram prompt produces stunning hand-drawn anatomy art — azed_ai · 2026-11-17
- REA-Video skill lets Claude reverse-engineer video editing styles; rival claims 20x speed at SSIM 0.96 — toolstelegraph · 2026-11-17
- H3 anime-to-photo conversion impresses: 16MP renders, 6-step LCM sampling — Loose-Journalist-555 · 2026-11-17
- Songwriter finishes AI music video from hospital bed with Suno and Runway — Kyrannio · 2026-11-17
- Ant Group open-sources Duo, a reciprocal world model where virtual agents decide when to speak — aigclink · 2026-11-17