Google Gemini Media API Examples

Compare and call Google image, video, multimodal video, and speech generation endpoints through Muapi's unified API. This repository focuses on Google's generative media workflows; it does not claim to expose the general Gemini chat API.

Related Projects

• Open-Generative-AI — open-source generative media application with multi-model workflows.
• Seedance-2-API — API examples for another video-generation family.
• Nano-Banana-2.1-API — Nano Banana 2.1 image generation and editing API guide with Python, JavaScript, and curl examples.

Overview

• Product page: Google API on Muapi
• API documentation: Muapi API docs
• Get an API key: Muapi access keys
• Website: muapi.ai
• Nano Banana: playground and API reference
• Imagen 4: playground and API reference
• Veo 3.1: playground and API reference
• Gemini Omni: playground and API reference
• Gemini TTS: playground and API reference

Muapi accepts a single authenticated REST interface for these products. Requests return a request_id; generation runs asynchronously and the result endpoint returns status and output URLs.

Endpoint coverage

Workflow	Endpoint	Muapi catalog price	Typical use
Nano Banana image generation	POST /api/v1/nano-banana-2	$0.06/request	Prompt-based image creation
Nano Banana image editing	POST /api/v1/nano-banana-2-edit	$0.06/request	Prompt-directed image edits
Imagen 4 standard	POST /api/v1/google-imagen4	$0.03/generation	General image generation
Imagen 4 Fast	POST /api/v1/google-imagen4-fast	$0.02/generation	Lower-latency image drafts
Imagen 4 Ultra	POST /api/v1/google-imagen4-ultra	$0.06/generation	Higher quality image generation
Veo 3.1 text-to-video	POST /api/v1/veo3.1-text-to-video	$2.50/request	Video generation from prompts
Veo 3.1 Fast text-to-video	POST /api/v1/veo3.1-fast-text-to-video	$0.60/request	Faster, lower-cost video generation
Gemini Omni text-to-video	POST /api/v1/gemini-omni-text-to-video	From $0.16	Video with synchronized generated audio
Gemini Omni 1.1 Flash text-to-video	POST /api/v1/gemini-omni-flash-1-1-text-to-video	From $0.10/sec	Multimodal prompt video with optional references
Gemini 3.8 Flash TTS	POST /api/v1/gemini-3-8-flash-tts	$0.015/1,000 characters	Speech and multi-speaker dialogue audio

Prices above are Muapi catalog prices observed in this repository's landing-page data, not Google's direct API prices. Request options can change billing; check the live model page and pricing before production use. Omni video prices vary by duration/resolution. The landing page lists Nano Banana 2; this repository also shows the original nano-banana endpoint in the quickstart because it has a particularly compact example.

Choosing an endpoint

Compare models against the same prompt and output requirements. Useful criteria:

1. Task fit: text-to-image, image editing, text-to-video, reference-conditioned video, or speech.
2. Output controls: aspect ratio, resolution, duration, number of images, and reference inputs supported by that endpoint.
3. Audio behavior: Veo and Gemini Omni endpoints differ in synchronized audio support and workflow; TTS returns speech audio from scripted dialogue.
4. Latency and cost: compare Fast variants against standard or Ultra tiers, and calculate video rates against the requested duration/resolution.
5. Input contract: check required image/video URLs, supported fields, and limits on the linked endpoint reference before submitting.
6. Output quality: evaluate prompt adherence, text rendering, motion, identity consistency, and voice quality on your own representative samples.

Quickstart

Set your key in the shell; do not commit it:

export MUAPI_API_KEY="your-api-key"

Submit an image request:

curl -X POST https://api.muapi.ai/api/v1/nano-banana \
  -H "x-api-key: $MUAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt":"A small glass greenhouse on a mossy hillside at sunrise","aspect_ratio":"1:1"}'

The response contains a request_id. Poll for completion:

curl "https://api.muapi.ai/api/v1/predictions/REQUEST_ID/result" \
  -H "x-api-key: $MUAPI_API_KEY"

The result response includes a status and, when complete, generated asset URLs in its output data. Use the exact endpoint-specific request fields shown in the model page before adapting these examples.

Examples

See examples/google_media.py for a dependency-free Python submit-and-poll flow with Nano Banana, Imagen 4, Veo, Gemini Omni, and Gemini TTS payloads. The example prints the final JSON response; it does not download generated assets.

Request patterns

All examples use https://api.muapi.ai/api/v1/{endpoint} and the x-api-key header. A successful submission returns a request identifier. Poll GET /api/v1/predictions/{request_id}/result until the request completes, then consume the returned asset URL(s). Generation is asynchronous, so allow time between polls and handle failed status responses in your application.

Field examples

• Image generation: prompt, with supported controls such as aspect_ratio, resolution, and num_images depending on model.
• Image editing: prompt plus the endpoint's image input field (for example, an image URL or list of image URLs).
• Video generation: prompt, with supported duration, resolution, and aspect_ratio fields. Gemini Omni may accept optional reference media on its expanded endpoint.
• TTS: speakers and ordered dialogue_turns; each dialogue turn refers to a declared speaker_id.

The schemas are not interchangeable between endpoints. Use the linked Muapi model references and landing page for current parameters, validation limits, and pricing.