ChatImageVideo

Model developer guide

Google

31 website model guides

Start by deciding whether you need an answer, an image or a video. Gemini analysis, Nano Banana image generation and Google video models have different contracts. This guide helps connect a practical brief to the right model guide without turning Google's wider platform into a list of promised app controls.

MultimodalImage generationVideo generation

Practical guidance

Where should I start?

Editorial starting points, not a performance ranking. Check current app access before choosing a model.

Analyze a difficult code or engineering brief

Evaluate Google’s flagship Argon reasoning model for complex coding and analysis. Its provider API has a unified 1M-token input/thinking/output budget, multimodal understanding and long-decode mechanisms.

Tradeoff: Input, thinking and output share one window. Longer responses need documented continuation or background execution, and a larger reasoning budget can reduce the space left for the visible answer.

Gemini 4 Argon guide

Synthesize supplied documents and media

Evaluate Gemini 3.8 Flash when your task combines written evidence with images, audio, video or PDFs. Google documents those input types and text output; it is an analysis choice rather than an image generator.

Tradeoff: The app may expose fewer upload types or tools than the provider. Give the model relevant excerpts and verify evidence locations instead of relying on the maximum context size.

Gemini 3.8 Flash guide

Develop a visual composition

Use an image-model guide when the deliverable is a visual rather than a written description. Google's image documentation covers creation and editing, while the audited app route here is text-led.

Tradeoff: Do not assume reference-image editing is available because Google's own API supports it. Check the exact model's app-control notes and review generated text and geometry carefully.

Nano Banana 2.1 guide

Plan a short moving scene

Veo 3.1 is documented for video with native audio and provider features such as frame-specific generation and extension. Compare that workflow with Gemini Omni rather than treating every Google video model as interchangeable.

Tradeoff: The audited Veo app route uses WaveSpeed, so Google's direct endpoint options do not establish what reaches that gateway.

VEO 3.1 guide

Technical details

What does the provider document?

Official documentation describes the named model or gateway—not every feature available in this app.

Gemini input and output budgets

The gemini-3.8-flash reference lists 1,048,576 input tokens and 65,536 output tokens. It accepts text, images, video, audio and PDF but produces text. These are API limits, not a promise of equivalent file sizes, upload support or unrestricted conversation history inside EZ Ai Assist.

Thinking and tools are explicit integration choices

Gemini 3.8 Flash supports low, medium and high thinking; minimal is unsupported. Google also lists search grounding, code execution, function calling and structured outputs. Each requires the appropriate integration. A model cannot search your files or execute a program merely because those capabilities appear in its documentation.

Media generation has separate endpoints

Google distinguishes Gemini image generation from video generation. Its video overview separates conversational Omni workflows from Veo-specific extension and frame controls. Input-image understanding in a chat model is not image editing, and understanding a supplied video is not generating a new one.

Complete website coverage

Explore all 31 model guides

Every existing catalog entry for this provider is linked below. Status labels come from the individual website guides; catalog coverage is not a guarantee of current app access. Historical and snapshot pages remain references, not extra active models.

Practical guidance

From a brief to a useful result

Original prompt examples, not generated-output demonstrations. They have not been generation-tested.

Workflow 1

Reconcile a workshop recording with its handout

Use a short recording, transcript or slide excerpts you are allowed to share. Attach media only if your current app session supports it; a transcript is a useful text-only fallback.

Compare this workshop transcript with the attached handout. List the steps that agree, instructions that conflict, and information present in only one source. Cite a timestamp, paragraph or page for each observation. Draft a short learner checklist using only supported instructions. Mark uncertainty rather than inventing missing steps.

Check the result: Verify every cited passage and inspect contradictions yourself. Check that the model has not turned a spoken suggestion into a confirmed requirement.

Read the Flash guide

Workflow 2

Create a clear editorial illustration

Choose the image model, then any aspect ratio and resolution actually available. Keep this a text-to-image brief; no reference editing or later layout software is implied by the request.

Create an original editorial illustration of a neighborhood library sharing tools as well as books. Show one librarian, one visitor and a clearly organized shelf of simple hand tools. Use a calm teal, rust and cream palette, clear silhouettes and generous empty space at the top for a headline added later. Include no lettering or real logos.

Check the result: Inspect hands, object count, tool shapes and headline space. Confirm the returned dimensions before placing the illustration in a page layout.

Read the Nano Banana 2.1 guide

Technical details

What changes inside EZ Ai Assist?

Google's official references support the provider capabilities below, including Argon’s shared context, thinking levels and long-decode contract. Image controls retain their October 8 owner-audit attribution and video controls retain their existing video-audit attribution. Those separate integration references are not live generation tests.

Google Gemini Developer API documentation

Argon has a distinct Google model identity

The stable model ID is gemini-4-argon. Google documents text, image, video, audio and PDF understanding with text output, plus Google Search grounding and compatible tool workflows. Its model guide separates the provider’s shared-context and long-decode specifications from session-level request settings.

Website catalog

The full list is website coverage

The model list includes active-looking entries alongside historical Imagen pages and preview references. It preserves their existing identities and URLs; it is not a live inventory of account access or a guarantee that a retained endpoint still accepts requests.

Owner image audit · October 8

Image controls are narrower than the API

The Google image routes are reported to send text only. Nano Banana 2.1 exposes aspect ratio and 1K/2K/4K resolution, while several other routes use provider defaults. The original Nano Banana uses Lovable AI Gateway. Preview mappings and migration dates have unresolved lifecycle differences noted in the model guides.

Owner video audit · not live-tested

Video settings depend on the route

The existing Veo guides report WaveSpeed routing and warn that the shared wrapper does not forward all advertised audio or resolution selections. Omni has different audited controls. For earlier Gemini chat routes other than Argon, thinking, tool settings and effective file limits remain unverified in this provider guide. Do not inherit one route's settings across the family.

Before you rely on a result

Limits, costs and tradeoffs

  • Large context is capacity, not guaranteed recall. Ask for evidence locations, split important reviews into checkable sections and validate omissions as well as errors.
  • Provider preview retirement, app migration timing and gateway availability are separate facts. A retained website URL does not silently select a newer model.
  • Direct Google API rates and gateway rates are not customer charges. Quality, latency and cost should be evaluated on the exact request path rather than copied from a provider headline.

Provider references and picker estimates are not EZ Ai Assist customer charges. Read the website pricing explanation.

Compare by task

Other paths to explore

OpenAI

Compare another text-and-image family using the same source material and acceptance checks.

xAI

Explore another set of separate chat, image and video contracts; compare the actual app settings.

Google Lyria

For music, use the dedicated Lyria family guide instead of assuming Gemini chat creates audio tracks.

Common questions

Google, without the assumptions

Can Gemini 3.8 Flash create an image or song?

Its cited model reference specifies text output, not image or audio generation. Select a dedicated image or music model for those deliverables. Multimodal input describes what a model can understand, not everything it can create.

Why keep old Google model pages?

They preserve existing links and explain lifecycle or app-mapping changes. Historical Imagen and preview references help migration planning; their presence does not certify current app availability.

Are Veo's audio and extension features app controls?

Not automatically. The provider documents them, but the owner audit describes a narrower gateway integration. Read the exact Veo guide before planning work that depends on audio settings, final frames or extension.

New to the terminology?

Official documentation

Sources & review notes

Highlighted provider contracts reviewed . Individual model guides carry their own review dates. Sources can change; these pages do not perform live health checks.

Google's official references support the provider capabilities below, including Argon’s shared context, thinking levels and long-decode contract. Image controls retain their October 8 owner-audit attribution and video controls retain their existing video-audit attribution. Those separate integration references are not live generation tests.