AI Video and Image Model Pricing 2026: Kling, Seedance, Veo, Wan, Sora, Nano Banana, GPT Image

Most "AI video generator pricing" pages compare monthly subscription tiers. That hides the number that actually decides your cost: what one second of video, or one image, costs at the model level. This page publishes those numbers.

Every figure below is the provider list price MakeInfluencer.ai pays to run the model through its inference partners, taken from the pricing manifest that drives the platform's own credit calculator. Each table notes the date the price was last verified. MakeInfluencer.ai converts these prices into credits, adds its margin, and shows the exact credit cost before every generation in the studio and through the preview-credits endpoint of the API.

If you only want the headline: a 5-second 1080p clip costs between roughly $0.19 (Alibaba Wan 2.6 Flash, image-to-video, silent) and $4.50 (ByteDance Seedance 2.5 at full quality). Images cost between half a cent (Z-Image Turbo) and $1.00 (OpenAI GPT Image 2.5 at max quality, 4K). Everything in between is a choice about resolution, audio and references, which is what the rest of this guide explains.

How AI video pricing actually works

Almost every frontier video model is priced per second of output. Four multipliers move the price:

• Resolution. 480p, 720p, 1080p and 4K are usually separate tiers, and the jump from 1080p to 4K is often 2x.
• Audio. Models that generate synchronized sound (Kling 3.0, Veo 3.1, Wan 2.6) charge more when audio is on, typically 1.5x to 2x.
• Reference images. Seedance 2.0 doubles its per-second price when you attach reference images because the model does more work to hold identity.
• Speed tier. Turbo and Flash variants (Seedance 2.5 Turbo, Wan 2.6 Flash, Seedance 2.0 Fast) trade some quality for a much lower price.

Image models are simpler: a price per image, sometimes tiered by resolution (1K, 2K, 4K) and by a quality setting (GPT Image 2.5 has five).

Subscription tools bundle all of this into "credits" without telling you the exchange rate. MakeInfluencer.ai uses credits too, but the credit cost is derived from the list prices below and shown before you generate, so you can always reason from the model price.

The quick answer: what a 5-second clip costs

The table below computes the provider list price of a single 5-second clip from the per-second rates in the sections that follow. Use it to rank models before you read the details.

Two things stand out. First, the cheapest way to get motion is image-to-video on a Flash tier: you lock the look in a still, then pay a fraction of the text-to-video rate to animate it. Second, the flagship ByteDance and OpenAI models are 5x to 20x more expensive per second than the mid-tier models, which is fine for hero shots and painful for drafts. Draft the still for free with the AI image generator, then test motion on a Flash tier in the image to video studio before spending on the expensive one.

Kling AI pricing (Kling 3.0 and motion control)

Kuaishou's Kling 3.0 is priced per second in two tiers, Standard and Pro, with a surcharge for generated sound. Prices verified 21 May 2026.

Kling's motion control mode, which copies the movement of a reference video onto your character, is priced separately and always includes the extra processing:

Kling 3.0 supports 3, 5, 10 and 15 second clips with start and end frames. For most creator content the Standard tier with sound off, then a separate lip-sync or music pass, is the cost-efficient path. See the Kling 3.0 guide and the Kling 3.0 model page for prompting and limits.

Is Kling AI free? Kuaishou's own consumer app has historically offered a small daily allowance of free credits with queue times and watermarks. Through any API, including MakeInfluencer.ai, Kling 3.0 is a paid model at the rates above. To test motion cheaply before paying Kling rates, draft the clip on Alibaba Wan 2.6 Flash in the image to video studio, then switch to Kling 3.0 when the concept works.

Seedance 2.5 pricing (and Seedance 2.0)

ByteDance's Seedance 2.5 is the most capable video model on the platform: 4 to 30 second clips, native synchronized audio, up to 30 reference images, and video edit, extend and talking avatar modes. It is also the most expensive per second at full quality. Prices verified 29 September 2026.

The Turbo tier is the value story: 1080p at $0.22 per second is a quarter of the full model's 1080p price, and for social clips the quality difference is small. A 15-second 1080p Turbo clip is $3.30; the same clip on full Seedance 2.5 is $13.50.

Seedance 2.0 remains available and is priced per second by resolution, with a 2x multiplier when you attach reference images. Prices verified 21 May 2026 (Fast tier 16 June 2026).

Practical rule: draft at 480p on Seedance 2.0 Fast (a 10-second draft is $1.00), then re-render the approved shot on Seedance 2.5 or 2.5 Turbo at the final resolution. The Seedance 2.5 API guide lists the five model IDs and request shapes.

Veo 3.1 pricing

Google's Veo 3.1 Fast is priced per second regardless of resolution, with a surcharge for generated audio. Prices verified 21 May 2026.

Veo 3.1 clips are 4, 6 or 8 seconds at 720p or 1080p, with start and end frames. Its native audio is among the most convincing for dialogue and ambience, so the audio surcharge is usually worth paying rather than adding sound afterwards. Prompting notes are in the Veo 3 prompting guide.

Wan 3.0, Wan 2.7 and Wan 2.6 pricing

Alibaba's Wan family covers the widest price range on the platform, from the cheapest image-to-video tier to a native-audio flagship. Wan 3.0 prices verified 29 September 2026; Wan 2.6 and 2.7 verified 21 May 2026.

Wan 2.6 Flash image-to-video is the cheapest way to test motion in the image to video studio: a 4-second silent 720p clip costs $0.10 at list price. Wan 2.7 accepts an audio file and times motion to it, which makes it the default engine for the AI music video generator. Wan 3.0 adds 2 to 30 second clips with native audio at a price that undercuts Seedance 2.5 Turbo at 720p. The Wan guide and the Wan 3.0 model page cover the differences.

Other video models: Gemini Omni, Happy Horse, Grok Imagine, Sora 2

Sora 2 pricing and availability. OpenAI retired the Sora 2 API on 24 September 2026, so the Sora 2 and Sora 2 Pro rates above are historical and the models are no longer selectable on MakeInfluencer.ai. Is Sora 2 free? The Sora consumer app offered free generation with daily limits in supported countries at launch; that is a separate product from the API and its terms change frequently, so check OpenAI's current help pages rather than any third-party summary, including this one. For Sora-style text-to-video with native audio today, Veo 3.1, Wan 3.0 and Seedance 2.5 are the closest alternatives, and the Sora 2 vs Kling and Sora 2 vs Veo comparisons explain the trade-offs.

Lip sync and talking avatar pricing

Lip sync is priced per second of the output video (or per second of the driving audio for Seedance's talking avatar mode).

A 30-second talking-head clip on InfiniteTalk at 720p is $1.80 at list price, which is the cheapest way to add speech to an AI presenter. The lip sync tool page explains inputs and limits.

AI image model pricing

Image prices are per image. Where a model has resolution or quality tiers, each tier is listed.

Nano Banana pricing (Google)

Is Nano Banana free? Google's Gemini app includes some free image generation with usage limits, but through APIs, including on MakeInfluencer.ai's Nano Banana generator page, the models are paid at the rates above. The free preview widget on that page runs our own open image model, not Nano Banana. See the Nano Banana Pro model page for capabilities.

GPT Image 2.5 pricing (OpenAI)

GPT Image 2.5 has five quality tiers and three resolutions. Text-to-image prices verified 29 September 2026.

Edits with GPT Image 2.5 Sunburst run on a slightly higher table (from $0.025 at low 1K to $1.015 at max 4K) and add $0.015 for each reference image beyond the first. The low tier at 1K is the cheapest way to draft with GPT Image 2.5 on the ChatGPT image generator page. For comparison, the previous GPT Image 2 model charged $0.06 at medium 1K and $0.22 at high 1K, so 2.5 is both sharper and cheaper at every tier. Model details are on the GPT Image 2.5 page.

Seedream, Dreamina, Qwen, Z-Image and Grok Imagine

Z-Image Turbo at half a cent per image is the draft engine: it is the model behind the free AI image generator and the free ad image generator, and MakeInfluencer.ai runs it on its own GPU capacity when available, which is what makes those tools free to offer.

Worked examples: what a real project costs

The numbers only matter in combination. Three common projects, at provider list price:

1. A week of AI influencer posts (7 images, 3 short videos). Seven Nano Banana Pro images at 2K ($0.14 each) come to $0.98. Three 5-second vertical clips on Kling 3.0 Standard, silent, cost $1.26. Total list price: about $2.24. Add sound on Kling for another $0.63 if the clips need ambience.

2. A UGC-style ad with a presenter. One Nano Banana Pro still ($0.14), one 15-second Seedance 2.5 Turbo clip at 1080p ($3.30), and 15 seconds of InfiniteTalk lip sync at 720p ($0.90). Total: about $4.34 for a finished presenter ad, before variations. Ten hook variations of the static alone add $1.40.

3. A 60-second AI music video. Twelve 5-second scenes on Wan 3.0 at 1080p ($1.00 each) come to $12.00, plus one Seedream 5.0 Pro character sheet ($0.09) and reference stills. Rendering the same scenes on full Seedance 2.5 at 1080p would cost $54.00, which is why the music video workflow drafts on Wan and reserves Seedance 2.5 for the two or three hero shots.

Because MakeInfluencer.ai adds a margin on top of list price, your credit cost will be higher than these figures. The proportions between models stay the same, and that is what should drive model choice.

Cost per finished second, not cost per generation

The per-second rates in this guide are what one generation costs. They are not what a usable second of footage costs, because almost nobody keeps the first take. The number that should drive your budget is the cost per finished second: everything you spent on a shot, including rejected drafts, divided by the seconds you actually publish.

Take a single 8-second hero shot on ByteDance Seedance 2.5 at 1080p. One generation is 8 × $0.90 = $7.20 at list price. Here are three ways to arrive at one keeper:

The same keeper costs anywhere from $0.99 to $3.60 per second depending on how you iterate. The model's headline price did not change; the workflow did.

There is a catch with the third row. A draft on a different model tells you whether the composition, framing and timing work, but it does not tell you exactly how Seedance 2.5 will move the subject. Drafting at 480p on the same model is the safer choice when motion or identity is the thing you are testing. Drafting on a cheaper model works best when you are testing the idea: does a slow push-in on a coffee cup read as a hook, does the outfit work against the background.

Two habits matter more than the model choice itself:

• Lock the first frame before you pay for motion. A still costs between $0.01 and $0.14 on the image models below. Generating ten candidate first frames and animating the best one is cheaper than generating ten text-to-video clips and hoping one lands.
• Track your keep rate. If one in three finals is usable, multiply your per-generation cost by three when you plan a month. Most creators underestimate this by half, which is why budgets built from list prices alone run out mid-month.

The iteration multiplier, by model

A rough guide from the shape of each model's pricing, assuming three drafts per final:

Seedance 2.5 Turbo has the worst draft ratio because its cheapest tier (720p) is almost the same price as its top tier (1080p). With Turbo, it is usually better to draft on Wan 3.0 at 480p, or skip drafting entirely and render at 1080p straight away, since the difference between a 720p draft and a 1080p final is only $0.16 on an 8-second clip.

Monthly budget scenarios: what 30 Reels a month costs

A creator posting one short vertical video a day needs about 30 finished clips a month. The table below prices three realistic production styles at provider list price, with the iteration built in. Credit totals use the MakeInfluencer.ai conversion explained in the next section.

Worked through, the lean style fits inside the Pro plan's 600,000 monthly credits and leaves about 188,000 credits for re-rolls. The balanced style needs the Studio plan (1,500,000 credits) plus a credit pack, or 24 Reels a month on Studio alone. The premium style is roughly two Studio plans' worth of credits, so most creators on that style mix it: Seedance 2.5 Turbo for the five or six Reels that carry a hook or a product, and Wan 3.0 or Wan 2.6 Flash for the rest.

A mixed month is where most accounts end up. Twenty-four Reels on the lean path (329,664 credits), six on the premium path (601,512 credits) and 30 extra GPT Image 2.5 medium stills for carousels (36,120 credits) come to 967,296 credits, which fits on Studio with room to spare for re-rolls. At list price that month is about $19.30.

Scenario: a small brand running paid UGC-style ads

Ads are the opposite of daily Reels: few finished assets, many variations. A typical test cycle is three presenter concepts, each with five hooks.

• Three character stills on Nano Banana Pro at 2K: 3 × $0.14 = $0.42.
• Fifteen 5-second hook clips on Seedance 2.5 Turbo at 1080p: 15 × $1.10 = $16.50.
• Three 20-second InfiniteTalk lip-sync bodies at 720p: 3 × $1.20 = $3.60.

That is $20.52 at list price for fifteen ad variants, or about $1.37 per variant. Because hooks are short and the body is shared, the lip-sync model does most of the talking at a fraction of the per-second price of a native-audio video model.

Scenario: an agency producing a 60-second product film each week

Four films a month, each built from twelve 5-second shots. Drafting every shot twice at 480p on Wan 3.0 (2 × $0.25) and finishing on Wan 3.0 at 1080p ($1.00) costs $1.50 per shot, $18.00 per film and $72.00 a month. Swapping the three hero shots in each film to full Seedance 2.5 at 1080p ($4.50 each instead of $1.00) adds $10.50 per film, taking the month to $114.00. This is the workload where credit packs and the Studio plan matter, and where the preview-credits endpoint earns its keep, because a single mis-set resolution on a batch of 48 shots is expensive.

How to spend less per finished video

• Draft low, finish high. Generate at 480p or on a Turbo tier until the prompt works, then re-render once at the final resolution.
• Prefer image-to-video for identity. Fix the character in a still, which costs cents, then animate on a Flash tier. Text-to-video re-rolls the whole scene every time.
• Turn audio on only for the final render. Audio surcharges of 1.5x to 2x add up across drafts.
• Watch reference-image multipliers. Seedance 2.0 doubles its price with references; Seedance 2.5 does not, which makes 2.5 Turbo the better choice for reference-heavy work.
• Use the free image preview for exploration. The free widget on the AI image generator and ad image pages runs our own open image model with a small daily limit and no account. It is enough to rough out a composition before you pay for GPT Image 2.5, Nano Banana or any video.
• Preview credits before you generate. Both the studio and the API show the exact cost up front, and the preview-credits endpoint lets you enforce a budget in code.

How MakeInfluencer.ai credits map to these prices

MakeInfluencer.ai sells credits, not seconds, so it is worth knowing exactly how a list price turns into a credit charge. There is one rule for every model on the platform: the credit cost of a generation is computed from the provider list price of that exact configuration (model, duration, resolution, audio, quality tier) using a fixed conversion of about 50,126 credits per $1 of list price, rounded up to the next whole credit. Nothing is priced by feel, and no model has a hidden flat fee. When a provider changes its price, the manifest is updated and the credit cost follows.

That gives a simple way to predict any charge: multiply the list price by 50,126. Some common outputs:

Two details affect the math. InfiniteTalk bills a 5-second minimum, and Kling motion control bills a 3-second minimum (5 seconds for Kling 2.6 Pro), so very short clips on those models cost the same as the minimum. Seedance 2.5 Talking Avatar bills per second of audio with a 4-second minimum.

What each plan buys

The current monthly plans, from the pricing page:

The "roughly equals" column spends the whole balance on one output type, which nobody does in practice; it is there to show scale. Credit packs and one-time bundles top up the same balance when a month runs over.

Plan tier also decides which settings you can use. 1080p and 4K video output, Seedance 2.5, Kling 3.0 Pro, Veo 3.1 and 4K image models require Pro or Studio. Creator covers the image lineup at standard resolutions, 720p video on the mid-tier models and Kling 2.6 motion control. If your budget scenario depends on 1080p finals, plan on Pro as the minimum.

What credits add on top of the list price

The list price is what the provider charges for inference. The credit price also covers failed-job refunds (a generation that errors out returns its credits automatically), storage of every output in your library, the character and reference system that keeps one AI persona consistent across models, and the studio, canvas, REST API and MCP connector that all draw on one balance. Whether that is worth it depends on how much of that wrapper you use. A developer who only needs raw inference for one model will usually pay less going straight to that model's provider; a creator switching between GPT Image 2.5, Seedance 2.5, Wan 3.0 and lip sync in one project is paying for not having to manage four accounts.

The REST API and the MCP connector charge the same credits as the studio, so there is no separate API price list. Developers can list models, call the preview-credits endpoint to get the exact charge for any input without spending anything, and generate through the endpoints documented in the API reference. The AI models hub has one page per model with capabilities and limits.

Frequently asked questions

How much does Kling AI cost per second?

Through the API, Kling 3.0 Standard is $0.084 per second silent and $0.126 with sound; Kling 3.0 Pro is $0.112 and $0.168. Motion control adds processing and ranges from $0.07 (Kling 2.6 Standard) to $0.168 (Kling 3.0 Pro) per second. Verified 21 May 2026.

How much does Seedance 2.5 cost?

Seedance 2.5 is $0.18 per second at 480p, $0.36 at 720p, $0.90 at 1080p and $1.80 at 4K. Seedance 2.5 Turbo is $0.20 at 720p and $0.22 at 1080p. Verified 29 September 2026.

Is Sora 2 still available, and what did it cost?

OpenAI retired the Sora 2 API on 24 September 2026. Before that, Sora 2 was $0.10 per second and Sora 2 Pro was $0.30 to $0.70 per second depending on mode and resolution. The consumer Sora app is a separate product with its own free limits.

What is the cheapest AI video model?

For image-to-video, Alibaba Wan 2.6 Flash at $0.025 per second (720p, silent) is the cheapest on this page. For text-to-video, Wan 3.0 at $0.05 per second at 480p and Grok Imagine Video at $0.055 per second are the lowest.

What is the cheapest AI image model?

Z-Image Turbo at $0.005 per image. Among frontier models, GPT Image 2.5 at low quality is $0.01 per 1K image, Dreamina V3 is $0.027 and Nano Banana is $0.039.

Does MakeInfluencer.ai charge the same as the provider list price?

No. The platform buys inference at the list prices shown here and charges a credit price that includes its margin. The exact credit cost is displayed before every generation, and the API has a preview-credits endpoint that returns it without charging.

Is there a free AI video generator without sign-up?

Not on MakeInfluencer.ai. Every video model is billed per second by its provider, so all video in the studio runs on credits, with the cost shown before you render. What is free without an account is an image preview on our own open image model, useful for drafting a start frame. Among other tools, the recurring free video allowances we know of, such as Google Flow and Kling's app, require an account.

How often are these prices updated?

Each table shows the date the price was last verified against the provider's public documentation. The platform's credit calculator reads from the same manifest, so a price change on the provider side is reflected in the studio when the manifest is updated.

Is Kling AI free?

Not through an API. Kling 3.0 is billed per second everywhere it is offered programmatically, at $0.084 to $0.168 per second at list price. Kuaishou's own consumer app has historically given new and returning users a small allowance of free credits with slower queues and watermarked output; the allowance changes often, so check Kling's current terms before relying on it. On MakeInfluencer.ai, Kling 3.0 Standard is available on every paid plan and a 5-second silent clip costs 21,053 credits.

Is Sora 2 free?

Sora 2 is no longer available through OpenAI's API, which OpenAI retired on 24 September 2026, so there is no free or paid API access to it. The consumer Sora app is a separate product with its own limits and regional availability. If you need Sora-style text-to-video with synchronized audio, Veo 3.1, Wan 3.0 and Seedance 2.5 are the practical replacements.

Is Nano Banana free?

Partly. Google's Gemini app includes limited free image generation, but MakeInfluencer.ai does not give Nano Banana generations away; the free preview on our Nano Banana page runs our own open image model instead. Through APIs, Nano Banana is $0.039 per image, Nano Banana 2 starts at $0.07 and Nano Banana Pro starts at $0.14 (verified 21 May 2026).

Is Seedance 2.5 free?

No. Seedance 2.5 is ByteDance's flagship video model and is paid wherever it is offered through an API, from $0.18 per second at 480p to $1.80 per second at 4K, with the Turbo tier at $0.20 to $0.22 per second (verified 29 September 2026). ByteDance's own consumer apps have run trial allowances on Seedance models at various points; those vary by region and are outside the scope of this page.

How much does GPT Image 2.5 cost?

From $0.01 per image (low quality, 1K) to $1.00 per image (max quality, 4K) for text-to-image, verified 29 September 2026. Medium quality at 1K, which is enough for most social posts, is $0.024. Edits run slightly higher and add $0.015 for each reference image beyond the first.

How much does Wan 3.0 cost?

Wan 3.0 is $0.05 per second at 480p, $0.10 at 720p and $0.20 at 1080p, with native audio included and clips from 2 to 30 seconds (verified 29 September 2026). Wan 3.0 Prime, the accelerated tier, is 1.5x those rates. A 10-second 720p clip costs $1.00 at list price.

How many credits does a 5-second AI video cost on MakeInfluencer.ai?

It depends on the model and settings, because credits are computed from list price at about 50,126 credits per dollar. A 5-second clip ranges from 6,266 credits (Wan 2.6 Flash image-to-video, 720p, silent) to 225,564 credits (Seedance 2.5 at 1080p). Kling 3.0 Standard silent is 21,053 and Wan 3.0 at 720p is 25,063. The studio shows the exact number before you confirm.

What does it cost to post an AI video every day for a month?

At list price, between about $8 and $60 a month for 30 short Reels, depending on the model and how many drafts you make. The lean path (a GPT Image 2.5 still animated on Wan 2.6 Flash) is about 412,000 MakeInfluencer.ai credits a month and fits on the Pro plan; a Wan 3.0 path with native audio needs Studio plus a top-up. The monthly scenarios section above breaks down each option.

Why do my credits run out faster than the per-second price suggests?

Almost always because of iteration. If you keep one clip in three, your real cost per finished second is three times the per-generation price. Drafting at 480p or on a Turbo or Flash tier, then rendering one final at full resolution, typically cuts the cost per finished second by half or more.

Prices are provider list prices in US dollars, verified on the dates shown, and exclude MakeInfluencer.ai's margin. Providers change prices without notice; treat this page as a reference, not a quote.