NANO BANANA 2 : mon SYSTEME pour générer des images PARFAITES

Item: NANO BANANA 2 : mon SYSTEME pour générer des images PARFAITES
Rating: 3
Author: PostSingularity Institute

NANO BANANA 2: My SYSTEM for generating PERFECT images

Title history

NANO BANANA 2 : mon SYSTEME pour générer des images PARFAITES — observed on 20261112-0057 UTC (YouTube)

by Thomas Berton · View the channel on PSI

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence

🎙 Thomas Berton 👥 48K 📅 March 12, 2026 ⏱ 13 min 👁 11K 📄 tutorial 🧭 2026-11-12

Available in: English (current) Français

Keywords

Nano Banana 2image generationprompt templateGoogle AIvoice dictation

Summary

In this tutorial, Thomas Berton introduces Nano Banana 2, officially known as Gemini 3.1 Flash Image Preview, a new image generation model by Google released on February 26, 2026. He highlights its superior speed, lower cost (6 cents per image), 4K resolution, and unprecedented face consistency. The video demonstrates the model’s capabilities through several tests: photorealistic portraits, cinematic scenes, text rendering, and thumbnail creation. The creator then shares his personal system: a JSON-structured prompt template that enhances results, which can be used in Claude or with a voice dictation tool (Speakly) for efficient prompting. He explains how to integrate the template into Claude projects and shows a workflow using a voice assistant to auto-generate prompts. The video concludes with a recap, free template access via a community school link, and invitations to engage with the channel. This content is a practical tutorial aimed at users interested in leveraging advanced AI image generation.

152 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers practical value by demonstrating the model’s capabilities and providing a ready-to-use prompt template, which is a significant time-saver for creators. The argumentation is built on live demonstrations and comparative tests (e.g., with and without the template), effectively showing the added value. However, the creator often relies on personal enthusiasm and statements like ‘it’s amazing’ without deeper technical explanations or external benchmarks. The reasoning is accessible but somewhat superficial; it focuses on ‘how to use’ rather than ‘why it works’.

Scientific Rigor, Source Quality, Title Accuracy

The creator includes official Google links in the description, such as the blog announcement and API documentation, which lend credibility. However, he also uses affiliate links (e.g., WaveSpeed AI, Speakly) and promotes his own services, being transparent about non-sponsorship but still blurring the line between education and marketing. The title accurately reflects the content, though the word ‘PERFECT’ is highly subjective. No external scientific sources or third-party evaluations are cited, limiting the rigour.

170 words

Title / Content Match

The title accurately reflects the content: the video explains a personal system for generating images using the new model, including a prompt template and voice integration.

Quality & Reliability

6/10

The video provides practical information about Google's new image generation model, Nano Banana 2 (Gemini 3.1 Flash Image Preview), with demonstrations and a free prompt template. The creator is transparent about unofficial aspects and includes official links, but claims are often hyperbolic and personal opinion, lacking rigorous scientific verification.

Chapters

Cited Sources

Gemini API - Image Generation Documentation — Official documentation for image generation with Gemini models, referenced for technical details.
Gemini 3.1 Flash Image Preview Model Page — Model card and specifications for the new image generation model.
Google Blog - Nano Banana 2 Announcement — Official Google blog post introducing Nano Banana 2 and its features.
Google Workspace Updates - Introducing Nano Banana 2 in Gemini — Announcement about the integration of Nano Banana 2 in the Gemini app.
AI Studio — Platform where the model is used for demonstrations, allowing access to Google's AI models.

Concurring Sources

Google Blog - Nano Banana 2 — Confirms model capabilities and release details presented in the video.
Gemini API Documentation — Provides technical specifications and usage guidelines aligning with the video's claims.

External References

Contribution & Novelties

This video contributes a practical, ready-to-use system for maximizing the output of a new AI image model. The original approach lies in combining a structured JSON prompt template with voice dictation, creating an efficient workflow for professional image generation. The creator also provides a free template, reducing the learning curve for newcomers.

For further exploration:

Prompt engineering — Core technique for optimizing AI outputs, highly relevant to the template approach.
Gemini (language model) — Background on Google’s Gemini family, of which Nano Banana 2 is part.
JSON — The data format used in the prompt template, important for structured prompting.

100 words

Radar Profile

The radar shows moderate scores across quantity, quality, technique, and reliability. The video excels in tutorial utility but lacks deep technical analysis and independent verification, resulting in a balanced but average profile.

Reliability 5/10