Nanobanana 2 is here!

Nanobanana 2 is here!

by AI Search

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 AI Search 👥 727K 📅 February 27, 2026 ⏱ 27 min 👁 79K 📄 review 🧭 2026-09-07
Available in: English (current) Français

Keywords

Nano Banana 2image generationAIGeminicomparison

Summary

The video is a comprehensive review of Google’s new image generation model, Nano Banana 2, which is based on Gemini 3.1 Flash. The creator conducts a series of challenging tests to evaluate the model’s capabilities, comparing it directly with the previous version, Nano Banana Pro. Tests include generating group selfies of celebrities, anime characters, emotion grids, Pokémon identification, Where’s Waldo scenes, endangered species knowledge, homework assistance, image-to-map conversions, spatial understanding, and data-to-chart conversion. The results show that Nano Banana 2 generally outperforms its predecessor in terms of speed, coherence, and prompt adherence, though it still struggles with certain tasks like generating accurate scientific illustrations and understanding complex spatial relationships. The video also covers the model’s specifications, such as its ability to maintain character resemblance for up to five characters and support for new aspect ratios. The creator highlights that Nano Banana 2 is available for free on Google’s platforms, including the Gemini app and AI Mode in Google Search. The video includes a sponsored segment for Higgsfield, a platform that integrates various AI models. Overall, the review provides a thorough, hands-on assessment of Nano Banana 2’s strengths and weaknesses.

189 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers substantial value by providing a practical, hands-on evaluation of Nano Banana 2, a cutting-edge AI image generator. The creator designs a series of creative and challenging tests that go beyond simple prompts, effectively probing the model’s world knowledge, prompt adherence, and editing capabilities. The argumentation is solid, as the creator systematically compares Nano Banana 2 with its predecessor, highlighting specific instances where one outperforms the other. The use of visual examples strengthens the claims, allowing viewers to see the results firsthand. However, the evaluation is subjective, relying on the creator’s personal judgment of ‘good’ or ‘bad’ outputs, and lacks quantitative metrics. The sponsored segment, while clearly marked, may introduce a subtle bias, though the creator maintains a critical stance throughout the main review.

Scientific Rigor, Source Quality, Title Accuracy

The video demonstrates a good level of scientific rigor, as the creator references official Google resources, including the blog post announcing Nano Banana 2 and the Gemini platforms. The tests are well-designed and cover a range of scenarios, from celebrity recognition to spatial reasoning, providing a comprehensive overview of the model’s capabilities. The title accurately reflects the content, which is a review of the new model. The creator also provides links to additional resources, such as the AI Search website and newsletter, which may offer further information. The analysis of viewer comments reveals a generally positive reception, with viewers appreciating the detailed testing and expressing excitement about the model’s potential, though some note concerns about censorship and refusal issues.

260 words

Title / Content Match

The title accurately reflects the content, which is a review of the new Nano Banana 2 image generator.

Quality & Reliability

8/10

The video provides a detailed, hands-on review of Nano Banana 2, with numerous comparative tests against the previous version. The creator demonstrates a good understanding of the model's capabilities and limitations, and includes references to official Google resources. However, the evaluation is subjective and lacks quantitative rigor, and the presence of a sponsored segment may introduce bias.

Chapters

Cited Sources

  • Nano Banana 2 official blog post — Official announcement and details about Nano Banana 2.
  • Gemini — Google's AI platform where Nano Banana 2 is available.
  • Google AI Studio — Google's AI development platform where Nano Banana 2 can be accessed.
  • Higgsfield — Sponsor's platform offering Nano Banana 2 and other AI models.
  • AI Search — Creator's website for AI tools and jobs.
  • AI Search Courses — Creator's educational courses on AI.
  • AI Search Newsletter — Creator's newsletter for AI updates.
  • AI Search Giveaway — Page for the RTX 5090 giveaway.
  • Nvidia RTX 5000 Ada — Creator's GPU equipment.
  • Dell Precision 5690 — Creator's laptop equipment.

Concurring Sources

Contribution & Novelties

The video provides an original, hands-on evaluation of Nano Banana 2, a newly released AI image generator, offering a detailed comparison with its predecessor. It introduces a series of creative and challenging tests that go beyond typical benchmarks, such as generating images of endangered species and converting floor plans to photos, which are not commonly seen in other reviews. The video also highlights the model’s new features, including support for extreme aspect ratios and its integration into Google’s AI Mode, providing practical insights for users.

Pour aller plus loin :

  • Gemini 3.1 Flash — Official blog post about the underlying model.
  • Image generation — Wikipedia article on image synthesis, relevant to the technology.
  • Prompt engineering — Wikipedia article on prompt engineering, relevant to the tests conducted.

126 words

Radar Profile

The radar profile shows a balanced performance across all dimensions, with slightly higher scores in information quantity and quality, reflecting the video's comprehensive and detailed review. The technical level is moderate, making it accessible to a broad audience, while reliability is solid due to the use of official sources and hands-on testing.

Reliability 7/10

💬 Positif. Sur les 30 commentaires analysés, le public est enthousiaste et apprécie la démonstration détaillée, bien que certains signalent des problèmes de censure et de refus du modèle.