Google has released the latest version of its AI image generator, Imagen 3 to the public. Google hinted at the release back in May at the Google I/O annual developer event.
Imagen 3 is expected to rival other AI models such as DALL-E 3, Grok-2 and Midjourney. Google has introduced a SynthID digital watermark in the new model to authenticate and trace the image origins. So images generated by the new Imagen model will come with its own SythnID digital watermark.
Imagen 3: New Features That Come With The New Version
Grok-2 has faced a few problems with its uncensored approach to image generation, but Google intends to avoid the same problem. Imagen will now feature a thorough and large-scale filtering system as well as data labelling. This is done so to prevent or reduce the amount of damaging content in the model’s database. This ultimately decreases the chances of generating output deemed harmful.
Furthermore, the tech giant has also improved the model’s comprehension of text prompts. It has also enhanced text rendering, a problem most AI generators face. Image quality has also been improved upon in Imagen 3. Tiny details such as a hand’s wrinkles can also be rendered.
There have been a few criticisms about the model’s restrictiveness. Some users have reported that Imagen did not allow them to generate simple inputs and denied nearly half of them, despite the inputs being generally harmless.
Imagen 3 is currently available in two versions, ImageFX and VertexAI.
Suggested:
Google’s Deepfake Battle: New Features Combatting Deepfakes.
Google Maps Update Brings Speedometer and Speed Limits to iOS and CarPlay.
Operations in Brazil on X Will Be Closed, Says Elon Musk