Hopp til hovedinnhold
Fredag 24. april 2026AI-nyheter, ferdig filtrert for ledere
SISTE:
DeepSeek åpner V4 Preview med 1M kontekst og API-kompatibilitetOpenAI lanserer GPT-5.5 for ChatGPT og CodexAnthropic og Amazon utvider AI-alliansen med 5 GW kapasitet og ny investeringDeepSeek åpner V4 Preview med 1M kontekst og API-kompatibilitetOpenAI lanserer GPT-5.5 for ChatGPT og CodexAnthropic og Amazon utvider AI-alliansen med 5 GW kapasitet og ny investering
Adobe Firefly Custom Models Opens to Everyone — Train AI on Your Own Art
AdobeFireflyAIDesignCreativityBrand

Adobe Firefly Custom Models Opens to Everyone — Train AI on Your Own Art

JH
Joachim Høgby
19. mars 202619. mars 20264 min lesingKilde:

Adobe has launched Firefly Custom Models into public beta. For the first time, creative professionals and brands can train their own AI image model on their own assets — and keep it completely private.

What Are Firefly Custom Models?

The service allows users to upload existing images, illustrations, or photographs to train a personalized AI model. The trained model will then generate new images that preserve:

  • Stroke weight and illustration style
  • Color palettes and lighting
  • Character design and recognizable visual features
  • Brand visual identity across campaigns

Models are private by default — the images you use for training are not used to train Adobe's general Firefly models.

From Private to Public Beta

Firefly Custom Models were announced as a private beta at Adobe Max in fall 2025. Now Adobe is opening it to everyone. Pricing has not been announced, but the service is available through Creative Cloud.

Practical for Brands and Design Teams

For companies with high content production volumes, this is a powerful tool: instead of starting from scratch on each project, teams can reuse one trained model across campaigns and channels. Adobe explicitly positions it as a solution for brand consistency at scale.

Adobe has long positioned Firefly as an "ethical alternative" to competitors, as the models are trained on licensed and public domain content. Custom Models extends this by giving users full control over their own training data.

Important Caveat

Adobe does not explicitly state that it prevents users from training models on works they don't own — which could create legal grey areas if the service is misused.

📬 Likte du denne?

AI-nyheter for ledere. Kuratert av en CIO som bygger det selv. Daglig i innboksen.