Pricing AI video honestly: what each minute costs

Every AI video tool talks about "minutes per month" without telling you what a minute actually costs them. We are going to break that down - because the gap between the marketing number and the real number tells you a lot about what kind of company you are buying from.

This post uses the same coffee video that appears throughout the rest of this set - a 5:27 reference render. Numbers are pulled from the actual render log.

What goes into a 5-minute video

Four cost buckets, no overhead.

Voice - $1.35

• ElevenLabs V3, Bradford voice
• 8,420 characters of script across 8 generation chunks
• Two re-rolls for swallowed consonants (not billed at premium tier; included)
• $0.16 per 1,000 characters at our Studio rate; round to $1.35 for the run

A 60-second video at the same script density costs about $0.27 in voice. A 15-minute long-form runs about $4.00.

Images - $0.40

• Gemini 2.5 Flash Image (Nano Banana), 40 hero frames
• 32 frames generated at the standard rate, 8 with reference-image input (slightly more expensive)
• 3 rerolls
• Roughly $0.40 end to end at current Gemini pricing

A short-form video at 12 frames costs ~$0.12. A 15-minute piece at 110 frames runs about $1.10.

Music - $0.30

• Three royalty-cleared stems pulled from a licensed library at $0.10 per cue
• One additional master-licence allocation amortised across the project at $0.00 (the library deal is unlimited within Studio)

This is the bucket most likely to look different on your side: if you bring your own music, it's $0. If you use the per-cue partner library, it's about $0.06–$0.10 per cue.

Render compute - $0.05

• 11 minutes 17 seconds of total render wall time
• ~4 vCPU, 8 GB RAM worker
• At our infra cost (~$0.27/hr for a render-class instance), that is $0.05 of compute

Render is the cheapest part. People assume it's expensive because they hear "AI" and think GPU; render is pure ffmpeg on CPU.

Total provider pass-through

$2.10 for a 5:27 finished MP4 at 1080p. That's the all-in raw cost.

What you actually pay

We charge by minute-of-output, not by line-item, because nobody wants to budget for "approximately how many V3 character rerolls will I need this month." Current plan economics:

Studio is the volume tier - at $0.44/min for finished output, it covers our provider costs ($0.42/min based on the breakdown above), our infra and orchestration costs ($0.10–0.12/min depending on workload), our staff, and our margin.

If you do the math: at Creator pricing, we have ~$0.21/min of margin after pass-through. At Studio, slightly less per-minute but more total volume.

We do not bill for retries

This is unusual in the category, so worth stating clearly:

• Voice re-rolls are not billed against your minutes. Bad take, regenerate, no charge.
• Image re-rolls are not billed against your minutes. Same.
• Failed renders are not billed. If pass 2 craters on a malformed fragment, you don't lose a minute.

Why: we eat the provider cost of the retry. It is small, and a pricing model that punishes iteration is a pricing model that produces bad videos. We would rather you do three takes of a chunk and pick the best one than ship the first take to save a credit.

Compared to alternatives

The honest comparison set, in three rows.

A human freelancer

A competent video editor + voiceover artist combo for a 5-minute scripted explainer:

• Script: $150–300 (writer)
• Voiceover: $200–500 (talent + studio)
• Stock images / illustrations: $80–200
• Editor: $400–800 (assembly, color, audio)
• Total: $830–1,800 for one video, with a turnaround measured in days

This is genuinely better quality than VideoCue on the upper end of the budget. It is also 400× more expensive and 50× slower. Use case: high-stakes brand work where the marginal $1,000 buys you something.

Runway / Sora / generic text-to-video tools

These price by second of generated footage, not by finished video. To produce a 5-minute narrative from Runway's longest-clip model:

• ~50 clips at 6 seconds each
• Roughly $0.50/clip on the volume tier
• Raw generation cost: $25 before any voice, music, or assembly
• Plus your time to assemble in an NLE

If your video is short (under 90 seconds) and visual-first, Runway is a reasonable choice. For 5-minute narrative pieces, the per-second pricing kills you, and you still have to assemble the video yourself.

A stock-video subscription

Storyblocks, Artgrid, or similar: $30–80/month for unlimited downloads. Add voiceover ($30/mo on ElevenLabs Creator, separately), assembly time (yours), and you're at $50–110/month for unlimited-ish videos, if you have the editing skill and time.

The cost case is real. The catch: stock footage cannot be POV-narrative coffee bean shots. You are limited to whatever footage exists. For explainers and B-roll-heavy content, this is fine; for narrative video that requires bespoke imagery, it doesn't work.

The honest verdict

VideoCue is cheapest only on the specific shape of work we are good at: long-form narrative with bespoke imagery. For short-form social, Runway is faster. For high-stakes brand, a human team is better. For B-roll-heavy explainers, a stock sub plus your own editing is cheaper.

We win the segment we built for. We are not going to claim more.

What gets cheaper from here

Two forces are working in your favor over the next 12 months.

Voice costs are dropping. ElevenLabs V3 character pricing is down about 35% over the last year. We expect another 20–30% over the next 12 months as competition increases. Voice is the biggest line item, so this matters.

Image costs are dropping faster. Gemini Flash Image pricing is down ~50% over the last year and the per-image quality is up. We expect the image bucket to halve again.

Render compute is flat. It is already a rounding error.

We pass these reductions through. When voice provider pricing drops, we either raise minute caps on existing plans or hold pricing and improve margin - never both at once, and never silently.

What we don't do

• No "credits." Credits obscure pricing. Minutes of output are honest.
• No surge pricing. Render queue depth doesn't affect what you pay; it affects how long you wait. Studio's priority queue is the only differentiation.
• No locked-in voices. The voice you use is the voice ElevenLabs charges us for. We pass through every voice in their public library, including the premium ones, at the same per-character rate.

A model for thinking about your own cost

If you are publishing one or two videos a month, Creator is right. If you are publishing 8+ a month for a channel or clients, Studio's math works. If you are an agency or developer integrating into your own product, the Studio API rate is built for you.

A reasonable framing: VideoCue is cheaper than a freelancer the moment you publish your second video. It is cheaper than Runway the moment your videos cross about 90 seconds. It is cheaper than a stock subscription if you need imagery that doesn't exist in any library.

What to read next

• Introducing VideoCue - the AI video tool we wanted to use - the why behind the product.
• Anatomy of a 5-minute AI narrative: "Your Life as One Coffee Bean" - the line-item breakdown above, in production context.
• From script to upload in under an hour: a creator's workflow - what the cost looks like across a real week of publishing.

See the pricing page for current plan numbers.