Gemini Omni Flash + Nano Banana 2 Lite: 2026 Pricing, Free Tier, vs Veo 3.1

On June 30, 2026, Google shipped two new Gemini multimodal models to developers via the Gemini API and Google AI Studio: Nano Banana 2 Lite (gemini-3.1-flash-lite-image), the fastest and cheapest text-to-image model in the Nano Banana family, and Gemini Omni Flash (gemini-omni-flash-preview), a multimodal video generation and conversational editing model that accepts text, image, and video inputs in a single prompt. Both models have non-trivial pricing implications for production teams:

• Nano Banana 2 Lite is priced at $0.034 per 1K image — a competitive per-image rate that puts it between BFL FLUX schnell ($0.003 per megapixel) and dedicated premium image APIs.
• Gemini Omni Flash is priced at $0.10 per second of video output — identical to Veo 3.1 Fast at 720p, but with conversational multi-turn editing that Veo 3.1 Fast does not support.

Both models are now in public preview as paid-tier endpoints on the Gemini API, and both are reachable through the new Interactions API for multi-turn image-to-video workflows. This guide covers the full pricing breakdown, the four Nano Banana tiers, how the two new models compare to Veo 3.1 Fast / FLUX schnell / Imagen 4, the chained image-then-video workflow that the Interactions API enables, and the known limitations of the public preview as of July 2026.

What Google shipped on June 30, 2026: two new Gemini models

The June 30 Gemini models announcement is the first dual release since the Gemini 3 launch in November 2025. The two models address different layers of the multimodal stack:

• Nano Banana 2 Lite is the image generation tier: a text-to-image model in the Nano Banana 2 family designed for high-throughput batch and real-time preview where latency and cost dominate. It is positioned as the recommended replacement for the legacy Nano Banana (gemini-2.5-flash-image), which Google is now deprecating for production use.
• Gemini Omni Flash is the video generation and editing tier: a multimodal model that accepts text, image, and video inputs in one prompt and produces up to 10-second 720p/1080p video outputs with conversational multi-turn editing. It is the first Gemini model that supports conversational video editing — you can issue follow-up prompts like "make the balloon blue" or "pan left", and the model preserves scene continuity while applying the edit.

The two models are not in the same pricing tier, and they do not substitute for one another. Nano Banana 2 Lite is an image generator (text and image in, image out). Gemini Omni Flash is a video generator and conversational editor (text, image, and video in; video out). The real leverage comes when you chain them using the new Interactions API, which lets you keep the image from Nano Banana 2 Lite as a reference for Omni Flash and stack up to three sequential edits in a session.

Both models are GA on the Gemini API and Google AI Studio as paid-tier public previews. Both are also surfacing in Google consumer products (AI Mode in Search, the Gemini app, Google Photos, NotebookLM, Google Flow, Google Ads) at varying latency tiers. For developers evaluating the new endpoints, the canonical test path is to use Google AI Studio for playground prototyping (no install, browser-based) and then drop into the Gemini API for production workloads with the same gemini-3.1-flash-lite-image and gemini-omni-flash-preview model IDs.

Nano Banana 2 Lite: $0.034 per 1K image is the headline

The Nano Banana 2 Lite pricing breaks into two layers: the headline per-image figure Google uses in the announcement, and the underlying per-token pricing on the Gemini API. Both matter.

The headline figure is the right benchmark for "what does a 1024x1024 image cost me", but the per-token rates are what you actually pay on the Gemini API. For a typical workflow where you generate one image from a short text prompt, the breakdown is roughly:

• Input tokens: ~50 tokens for a one-sentence prompt at the standard Gemini tokenization = $0.0000125 per call
• Output tokens: 1120 tokens for a 1024x1024 image (Gemini's image-to-token count) = $0.00168 per image for the text/thinking portion
• Image output token equivalent: the rest of the $0.034 / image, billed as image-output tokens
• Total per image: ~$0.034

For a workload generating 1,000 images per day at 1024x1024, the cost is roughly $34/day. At 10,000 images per day, it is $340/day. At 100,000 images per day (a serious production workload for ad creative, e-commerce catalog expansion, or social content), the cost is $3,400/day, or ~$102K/month. That is high enough that the per-image rate is the binding constraint, not the inference latency, which is roughly 4 seconds per image at 1024x1024.

If you are coming from BFL FLUX.1 [schnell] at $0.003 per megapixel (~$0.003 per 1024x1024 image), Nano Banana 2 Lite is roughly 11x more expensive per image. If you are coming from OpenAI DALL-E 3 at $0.04 per 1024x1024 image, Nano Banana 2 Lite is roughly 15% cheaper. If you are coming from Replicate's hosted FLUX schnell at ~$0.004 per megapixel, Nano Banana 2 Lite is roughly 8.5x more expensive. The Nano Banana 2 Lite's pricing edge is not the lowest in the market — it is competitive, but the value is in the integration: native on the Gemini API alongside text reasoning, with multi-modal chaining through the Interactions API.

Gemini Omni Flash: $0.10 per second of video output

The Gemini Omni Flash pricing is split across input tokens, output text/thinking tokens, and output video tokens. The headline $0.10 per second is the effective rate for the video output portion translated to a per-second-of-video basis; the underlying token pricing is more granular.

For a typical 10-second 720p generation, the cost is roughly $1.00 in video output tokens, plus a small amount for the prompt input tokens (typically ~$0.001 for a short text prompt) and the implicit text/thinking tokens the model produces to plan the video (~$0.001-0.01 depending on prompt complexity). The total cost for a one-shot 10-second video is approximately $1.01-1.02 per generation.

The conversational editing capability changes the cost calculus. A multi-turn workflow that generates one 10-second clip and then applies 2-3 conversational edits (the Interactions API allows up to 3 sequential edits per session) does not regenerate the entire video from scratch on each edit. Omni Flash applies the edit to the existing clip, which means each edit costs roughly $0.20-0.40 in video output tokens rather than the full $1.00. For a workflow that needs "make the lighting warmer" or "pan the camera left" style follow-ups, the multi-turn editing is a substantial cost saving.

For a workload generating 100 video clips per day at 10 seconds each, the one-shot cost is $101/day. With conversational editing saving roughly 60% on follow-up edits, a typical multi-turn workload (1 initial gen + 2 edits per clip) costs roughly $46-66/day, depending on the size of the follow-up edits. At 1,000 clips per day, the cost is roughly $460-660/day, or ~$14K-20K/month. This is in the range where video generation becomes a viable line item for production teams, but not so cheap that you can ignore the budget.

The four Nano Banana tiers: Lite, 2, Pro, legacy

Nano Banana is the consumer-facing name for Google's Gemini native image generation model family. As of July 2026, the family has four active tiers, and the right choice depends on whether you are optimizing for speed, cost, or quality.

The choice between Nano Banana 2 Lite and Nano Banana 2 is the core decision for most teams. The Lite tier is roughly 25% cheaper per image ($0.034 vs $0.045) and roughly 3x faster (~4 sec vs ~12-15 sec). The trade-off is image quality: Lite produces "good enough" quality for thumbnails, product preview, and ideation, while 2 produces production-quality output for hero shots, advertising, and e-commerce model photography. The two are not substitutes — they target different points on the cost-quality curve.

Nano Banana Pro is the high-end tier. The pricing structure is unusual: image input tokens cost $2.00 per 1M tokens (same as Gemini 3.1 Pro text input), but the per-image rate for input is $0.0011, and the output image tokens cost $120 per 1M tokens. Pro is for workflows where prompt accuracy and reasoning matter more than speed or cost — design briefs with multiple constraints, brand-guideline-heavy campaigns, or images that need to compose with specific elements that Lite and 2 cannot handle reliably.

The legacy Nano Banana (gemini-2.5-flash-image) is still available but deprecated for new builds. Google explicitly recommends Nano Banana 2 Lite as the drop-in replacement, and the legacy model will likely sunset within 6-12 months. For a new application, do not start with the legacy model.

Nano Banana 2 Lite vs Imagen 4, FLUX schnell, SD 3.5

The image generation API market in July 2026 has five serious contenders in the speed/cost tier that Nano Banana 2 Lite competes in. Each has a different pricing model and a different integration story.

For a workflow that uses Gemini text models alongside image generation (the Interactions API chaining story), Nano Banana 2 Lite is the lower-friction choice. For a pure-image workload that needs the absolute lowest per-image cost, BFL FLUX.1 [schnell] at $0.003 per megapixel is still the cheapest in the market by roughly 11x. For a workflow that needs conditioning support (ControlNet-style structural inputs), Stability AI SD 3.5 is the only mainstream contender. For a workflow that needs the strongest photorealism, Imagen 4 Ultra is the right call despite the higher price.

The Nano Banana 2 Lite's pricing edge is the integration: it is the only one of these five that is native on the Gemini API alongside text reasoning, which means the Interactions API can chain image and text (and video with Omni Flash) in a single session without exposing separate API keys or managing cross-provider routing. For a team that is already paying the Gemini API bill and does not want to add a second image generation vendor, Nano Banana 2 Lite is the path of least resistance.

Gemini Omni Flash vs Veo 3.1 Fast, Runway Gen-4, Kling 2.1

The video generation API market in 2026 is thinner than the image generation market, but the four serious contenders are differentiated enough that the choice depends on the workflow.

Gemini Omni Flash and Veo 3.1 Fast are at the same per-second rate ($0.10 at 720p), so the differentiation is the conversational editing. If your workflow only needs one-shot generation and you do not need to refine a clip with follow-up prompts, Veo 3.1 Fast on the Gemini API is the simpler path (it has been GA longer, more tooling exists, and the documentation is more complete). If your workflow needs iterative refinement ("make the lighting warmer", "change the color of the shirt"), Omni Flash's multi-turn editing is the differentiator, and the per-second cost saving on follow-up edits makes it cheaper even at the same headline rate.

For a workflow that needs character motion transfer (e.g., a person doing a specific action that you want to apply to a new scene), Runway Gen-4 with the Act-Two feature is the only mainstream option as of July 2026. Omni Flash does not yet support motion reference inputs. For a workflow that needs the strongest motion realism (drone footage, sports, fast camera movement), Kling 2.1 is competitive with Gemini Omni Flash on quality but cheaper on price for one-shot generation. The right choice depends on whether you need conversational editing, motion transfer, or maximum realism.

The Interactions API: chaining image and video in one call

The Interactions API is the multi-turn orchestration layer on the Gemini API that maintains session history and context across model calls. It is the path of least resistance for any workflow that needs to chain Nano Banana 2 Lite (image gen) and Gemini Omni Flash (video gen) in a single session.

A typical image-to-video workflow using the Interactions API looks like this:

1. The agent calls Nano Banana 2 Lite with a text prompt and gets back an image URL.
2. The agent passes the image URL as a reference input to Gemini Omni Flash with a video-generation prompt ("animate the image with the camera panning right").
3. The agent receives the video output URL.
4. The agent can issue up to 2 additional follow-up edits in the same session (the Interactions API allows up to 3 sequential edits total).

The cost breakdown for this workflow:

• Nano Banana 2 Lite initial image: ~$0.034
• Gemini Omni Flash initial 10-sec video: ~$1.01
• Two conversational edits: ~$0.20-0.40 each = ~$0.40-0.80
• Total per image-then-video workflow: ~$1.44-1.84

For a workload generating 100 such workflows per day, the cost is roughly $144-184/day, or ~$4.4K-5.6K/month. This is the right cost tier for boutique creative production (advertising agencies, marketing teams, content studios). It is not cheap enough for high-volume social content generation (where you would need 10K+ workflows per day), but it is priced for the kind of iterative creative work that previously required a video editing team.

The Interactions API also supports the demo apps Google shipped alongside the announcement: Anywhere (transport a selfie to iconic landmarks using Nano Banana 2 Lite, then animate with Omni Flash), Space Lift (interior design via Nano Banana 2 Lite, cinematic showcase via Omni Flash), and Omni product studio (convert static images to cinematic e-commerce videos). These demos are remixable on the Google AI Studio side and serve as canonical examples of the chaining pattern.

The five known limitations of the public preview

Both models are in public preview as of July 2026, and there are five known limitations worth flagging before you commit to a production workload:

1. Omni Flash is limited to 10-second clips. Google has stated that longer durations are coming but has not published a timeline. For a workload that needs 30-second or 60-second clips, you would need to chain multiple Omni Flash outputs and stitch them together, which adds complexity and cost.
2. Audio references and scene extension are not supported. The blog post explicitly notes that "uploading audio references and scene extension is not yet supported in the Gemini API for this model." For a workflow that needs sound design or music alignment, you would need to handle audio in a separate pipeline (e.g., ElevenLabs for music and SFX, then ffmpeg to mux with the video).
3. Video references up to 3 seconds are accepted but not processed correctly. The API schema accepts video reference inputs up to 3 seconds in duration, but the model does not actually use them. Until this is fixed, video-reference-driven workflows should not rely on Omni Flash.
4. Character consistency has limitations across scene changes and panning. Google acknowledges that character consistency is imperfect when the scene changes or the camera pans. For workflows that need consistent character rendering across multiple angles (fashion, e-commerce, advertising), you should plan for regeneration rather than expecting perfect continuity.
5. No free tier during the preview. Unlike most Gemini API public previews, there are no free credits for Omni Flash. You are paying full price ($0.10/sec) from the first call. Plan your prototyping budget accordingly.

Nano Banana 2 Lite has fewer preview limitations — it is essentially a production-ready endpoint — but the legacy Nano Banana (gemini-2.5-flash-image) is being deprecated, and Google has flagged that Lite is the recommended replacement. The two models run side-by-side for now, but new builds should target Lite.

A 5-minute test plan for engineering teams

If you are evaluating the two new models for a production workload, the canonical test plan is five steps and can be completed in Google AI Studio with no install:

1. Open Google AI Studio and create a new prompt in the Gemini API playground.
2. Test Nano Banana 2 Lite with your most common prompt pattern. Measure the 4-second latency claim, the per-image cost from the response metadata, and the visual quality on your actual workload (not the canned demo prompts).
3. Test Gemini Omni Flash with a 10-second video prompt that matches your production pattern. Verify the conversational editing by issuing 1-2 follow-up prompts in the same session and confirming the video updates without regenerating from scratch.
4. Test the Interactions API chaining by calling Nano Banana 2 Lite first, then passing the resulting image to Omni Flash as a reference input. Verify the session context is preserved and the total cost matches the expected $1.44-1.84 per workflow.
5. Stress-test the rate limits by running 10-20 concurrent calls. The preview rate limits are stricter than the stable Gemini API rate limits, and you want to know your throughput ceiling before committing to a production rollout.

Once you have validated the workflow in Google AI Studio, drop into the Gemini API for production. The model IDs are gemini-3.1-flash-lite-image for Nano Banana 2 Lite and gemini-omni-flash-preview for Gemini Omni Flash. The Interactions API is reachable via the same Gemini API endpoint with the interactions resource path. The Python and Node.js SDKs wrap both endpoints with the same authentication surface as the rest of the Gemini API.

FAQ

What is Gemini Omni Flash?

Gemini Omni Flash (gemini-omni-flash-preview) is Google's multimodal video generation and conversational editing model, in public preview as of June 30, 2026. It accepts text, image, and video inputs in a single prompt and produces 10-second 720p/1080p video outputs with conversational editing across multiple turns. Pricing is $0.10 per second of video output (matching Veo 3.1 Fast at 720p). It is available in the Gemini API and Google AI Studio as a paid-tier preview, with no free tier during the preview window.

What is Nano Banana 2 Lite?

Nano Banana 2 Lite (gemini-3.1-flash-lite-image) is Google's fastest, most cost-efficient native image generation model, released June 30, 2026 as part of the Nano Banana 2 family. It generates text-to-image outputs in roughly 4 seconds at 1024x1024 and is priced at $0.034 per 1K image (the per-token equivalent is $0.25 input / $1.50 output per 1M tokens). Google recommends it as a drop-in replacement for the legacy Nano Banana (gemini-2.5-flash-image) and targets high-throughput batch and real-time preview workflows.

How much does Gemini Omni Flash cost?

Gemini Omni Flash is priced at $0.10 per second of video output. For text/image/video/audio input tokens, the price is $1.50 per 1M tokens. For output tokens (text and thinking), it is $9.00 per 1M tokens, and for output video tokens, it is $17.50 per 1M tokens. The $0.10/sec figure Google cites refers specifically to the video output tokens translated at the effective rate on a 1024x1024 / 24fps-equivalent basis, matching Veo 3.1 Fast at 720p.

How much does Nano Banana 2 Lite cost?

Nano Banana 2 Lite is priced at $0.034 per 1K image (the headline figure Google uses in the announcement). The token-equivalent pricing is $0.25 per 1M input tokens (text/image/video), $1.50 per 1M output tokens (text and thinking tokens), and the effective per-image price for a 1024x1024 image at 1120 tokens is $0.0336. The model has no free tier during the public preview; only the paid Gemini API tier is available.

Is there a Gemini Omni Flash free tier?

No. Gemini Omni Flash is only available on the paid tier of the Gemini API as a public preview. Input tokens are $1.50 per 1M, output text/thinking tokens are $9.00 per 1M, and output video tokens are $17.50 per 1M. There is no free trial credits path for Omni Flash specifically, though the Gemini API free tier covers Nano Banana 2 Lite text/image generation up to standard rate limits.

Which Nano Banana model should I use?

Use Nano Banana 2 Lite for high-throughput batch and real-time preview where speed and cost are the primary constraints ($0.034 per 1K image, 4-second latency). Use Nano Banana 2 (gemini-3.1-flash-image) as the generalist workhorse for most production workflows ($0.045 per 0.5K image, equivalent to $0.09 per 1024x1024 image, 12-15 second latency). Use Nano Banana Pro (gemini-3-pro-image) for complex professional use cases where prompt accuracy and reasoning matter more than speed ($0.0011 per image for text/image input tokens, $120 per 1M output image tokens). The legacy Nano Banana (gemini-2.5-flash-image) is still available but no longer recommended.

How does Nano Banana 2 Lite compare to BFL FLUX schnell?

BFL FLUX.1 [schnell] is priced at $0.003 per megapixel (~$0.003 per 1024x1024 image) and is roughly 10x cheaper than Nano Banana 2 Lite at $0.034 per 1K image. However, Nano Banana 2 Lite is delivered as a native endpoint on the Gemini API alongside text reasoning (no separate image pipeline needed) and produces outputs in ~4 seconds vs FLUX schnell at ~2-3 seconds. For a workflow that uses Gemini text models alongside image generation, Nano Banana 2 Lite is the lower-friction choice. For a pure-image workload that needs the absolute lowest per-image cost, FLUX schnell is still the cheapest in the market.

Is Gemini Omni Flash cheaper than Veo 3.1 Fast?

Gemini Omni Flash is priced at $0.10 per second of video output, which is the same as Veo 3.1 Fast at 720p resolution. Veo 3.1 Fast is $0.10/sec at 720p, $0.12/sec at 1080p, and $0.30/sec at 4k. Omni Flash's $0.10/sec figure is the 720p equivalent. The differentiator is that Omni Flash supports conversational multi-turn editing (you can refine the same clip with natural language follow-ups), while Veo 3.1 Fast is one-shot generation only. For a workflow that needs iterative video editing, Omni Flash is the cheaper path even at the same per-second rate.

What is the Gemini Interactions API?

The Interactions API is the multi-turn orchestration layer on the Gemini API that maintains session history and context across model calls. When you chain Nano Banana 2 Lite (image gen) and Gemini Omni Flash (video gen) using the Interactions API, you can pass an image from one as a reference to the other, and you can stack up to three sequential edits per session. The Interactions API is the path of least resistance for any workflow that needs image-to-video animation or iterative refinement across modalities.

Are Nano Banana 2 Lite and Gemini Omni Flash open source?

No. Neither Nano Banana 2 Lite nor Gemini Omni Flash has open weights as of July 2026. The Gemini API and Google AI Studio are the only commercial paths. Some earlier Nano Banana variants (the legacy gemini-2.5-flash-image) appear in third-party redistributors such as Replicate and fal.ai, but those are routed through Google's API with a 5-15% aggregator markup, not standalone open weights. For a self-hostable image generation path with open weights, BFL FLUX.1 [schnell] (Apache 2.0) remains the canonical option.

---

Reviewed against: Google blog announcement "Start building with Nano Banana 2 Lite and Gemini Omni Flash" (June 30, 2026), Gemini API pricing page (ai.google.dev/gemini-api/docs/pricing, last modified June 30, 2026), Gemini API image generation docs, Gemini Omni model page on DeepMind, Gemini API rate limits docs (last updated June 2026).

Disclosure: This article contains affiliate links. If you sign up through these links, we may earn a commission at no extra cost to you. Our reviews remain independent.