What a 10-minute faceless video actually costs.
Direct spend on a 10-minute faceless video runs from under $5 to over $100 depending on the route. Modeled at documentary-grade iteration on published July 2026 API prices, DIY with a Claude-class API lands around $45 to $90 in model spend, and a GPT-class API around $47 to $95, plus roughly $3 to $6 of voiceover either way. VidRush states $1.93 to $2.72 per finished minute on its published plans. CTRmaxxing works out to $1.15 to $1.42 per finished minute on current paid plans, about $11.54 to $14.21 per 10-minute video. The bigger gap is time: DIY takes roughly 10 to 30 hands-on hours, CTRmaxxing about 20 minutes of active work.
The cost comparison
| Route | Cost per 10-minute video | Per finished minute | Basis |
|---|---|---|---|
| DIY, Claude-class API (documentary-grade) | ~$45 to $90 model spend, plus ~$3 to $6 voiceover (our estimate) | ~$4.80 to $9.60 (modeled) | Modeled below on Claude Opus 4.8 rates ($5 in / $25 out per million tokens, as of July 2026). The flagship Fable 5 tier roughly doubles it. |
| DIY, GPT-class API (documentary-grade) | ~$47 to $95 model spend, plus ~$3 to $6 voiceover (our estimate) | ~$5.00 to $10.00 (modeled) | Modeled below on gpt-5.6-sol rates ($5 in / $30 out per million tokens, as of July 2026). The mid-tier gpt-5.6-terra is roughly half. |
| DIY, bare-minimum (small models) | Under ~$5 marginal cost (our estimate) | Under ~$0.50 (modeled) | Single research pass, two drafts on small models, one voiceover take. Quality is a single-pass draft; subscription floors and your hours are the real cost. |
| VidRush | ~$20 to $28, our division of their listed monthly prices by included minutes (monthly billing) | $1.93 to $2.72, their published figure | Plans of $99 to $639 per month including 36 to 327 video minutes, as of July 2026 (vidrush.ai/pricing). |
| CTRmaxxing | About $11.54 to $14.21, depending on plan | $1.15 to $1.42 | Computed at page render from our current plan config, the same source /pricing renders from. Per-plan breakdown below. |
DIY figures are good-faith estimates modeled on published API pricing as of July 2026; the full assumptions and arithmetic are published below so you can audit and re-run them. Stock footage is assumed from free libraries at $0; paid footage libraries are a monthly subscription on top. VidRush per-minute figures are their published numbers; the per-video figure is our own division of their listed prices and is labeled as such.
The time comparison
| Route | Hands-on time | Where the time goes |
|---|---|---|
| DIY, documentary-grade | Roughly 10 to 30 hours (our estimate) | Guiding research (1 to 3 h), steering script drafts (2 to 6 h), sourcing footage (2 to 6 h), editing and assembly (4 to 12 h), plus ideation, voiceover QC, and packaging. |
| DIY, bare-minimum | Roughly 4 to 8 hours (our estimate) | Editing dominates even a minimal build; the script and voice steps compress, the assembly does not. |
| VidRush | Not benchmarked | They do not publish a hands-on time figure and we have not measured one, so this benchmark does not assert one. |
| CTRmaxxing | About 20 minutes of active work | Type the idea, review the generated package (script, titles, description, thumbnail), start the render. Rendering finishes on its own, so elapsed time is longer than hands-on time, but you are not at the keyboard for it. |
Time figures are estimates from a typical solo workflow; skilled editors compress the top end of the DIY ranges.
The modeled arithmetic
Estimates based on published API pricing as of July 2026. Actual workloads vary widely; the arithmetic and assumptions are shown so you can audit and re-run them. Common assumptions across the DIY routes:
- A 10-minute video with a roughly 1,500-word script (about 150 words per minute of narration).
- "Documentary-grade" means many iterations: deep research passes, multiple full script drafts, fact-check and line-edit passes, and packaging with retries.
- Agentic workflows resend context each turn, so input tokens dominate. Prompt caching reduces this; the ranges assume only partial caching.
- Voiceover via ElevenLabs (costed below). Stock footage assumed from free libraries at $0.
Route A: Claude-class API, documentary-grade
Modeled on Claude Opus 4.8 at $5 input / $25 output per million tokens, as of July 2026:
| Step | Assumption | Input tokens | Output tokens | Input cost | Output cost |
|---|---|---|---|---|---|
| Research | 12 deep passes, ~400k in + 20k out each (context resent; priced at the full uncached input rate, so caching can only bring the real bill down) | 4.8M | 240k | $24.00 | $6.00 |
| Script drafts | 8 full drafts and revisions, ~150k in + 12k out each | 1.2M | ~100k | $6.00 | $2.50 |
| Fact-check + line edits | 6 passes, ~100k in + 8k out each | 600k | 48k | $3.00 | $1.20 |
| Packaging | Titles, description, thumbnail prompts, retries: ~300k in + 30k out total | 300k | 30k | $1.50 | $0.75 |
Model subtotal: about $45. Real projects hit dead ends and retries (a 1x to 2x multiplier), so this working style lands around $45 to $90 in model spend per video. Run the same workload on Claude Fable 5 ($10 / $50 per million tokens) and it roughly doubles to $90 to $180. Add voiceover (about $3 to $6, below) and any paid footage. A heavy, flagship-model, many-iteration Claude route lands in the tens to low hundreds of dollars per video, before your own time.
Route B: GPT-class API, documentary-grade
Same token workload as Route A on gpt-5.6-sol ($5 input / $30 output per million tokens, as of July 2026): input cost is identical at about $34.50 across all steps, output about 20 percent higher (about $12.50 versus $10.45). Subtotal about $47; with the same 1x to 2x retry multiplier, about $47 to $95 in model spend. On the mid-tier gpt-5.6-terra ($2.50 / $15) it is roughly half that, about $24 to $48. Same voiceover and footage adders.
Route C: bare-minimum cheap route (the honesty check)
Small models (Claude Haiku 4.5 at $1 / $5, or gpt-5.4-mini at $0.75 / $4.50 per million tokens), one research pass, two drafts, no deep iteration: about 150k input + 40k output tokens is well under $1 of model spend. One voiceover take is about $1.50 to $2 of ElevenLabs plan credits (or the free tier). Free stock footage, free editor. Marginal cost: under about $5 per video. The honest caveats: quality is a single-pass draft, you still pay subscription floors (for example ElevenLabs Creator at $22 per month), and above all your own hours. Cheap in cash is not cheap in time, and single-pass scripts are exactly the "AI scripts that get 200 views" problem.
Voiceover arithmetic (both DIY routes)
ElevenLabs prices text to speech at 1 credit per character on its standard model (as of July 2026, elevenlabs.io/pricing). Our computation: a 10-minute script at ~150 words per minute is ~1,500 words, roughly 8,000 to 9,500 characters, so ~8k to 10k credits per clean read. On the Creator plan ($22 per month for 121k credits) that is about $1.50 to $1.80 of plan credits per read; two to three regenerated takes lands around $3 to $6. The free tier's 10k credits roughly cover one 10-minute read. OpenAI also offers dedicated text-to-speech models (TTS-1 and TTS-1 HD) through its API speech endpoint; we anchor both DIY routes on ElevenLabs as this benchmark's chosen narration assumption because its published per-character plan pricing makes the arithmetic easiest to audit.
Sensitivity notes
- Token counts are the big lever: halve the research volume and Route A's floor halves too.
- Prompt caching (cache reads at 0.1x input price) and batch APIs (a 50 percent discount) can substantially cut these numbers for a disciplined pipeline; the ranges assume only partial use.
- Current frontier models tokenize roughly 30 percent more tokens for the same text than older models (per Anthropic's pricing docs); these rates already reflect that, so older per-token comparisons undercount.
VidRush, verified
All figures below are from VidRush's live public pages as of July 2026 (vidrush.ai/pricing and vidrush.ai).
| Plan | Monthly | Yearly (per month) | Video minutes | Videos / month | Team seats |
|---|---|---|---|---|---|
| Starter | $99 | $79.20 | 36 | up to 6 | 0 |
| Creator | $239 | $191.20 | 109 | up to 18 | 1 |
| Pro | $639 | $511.20 | 327 | up to 56 | 3 |
Their site-wide figure: "Between $1.93 and $2.72 per finished minute, depending on your plan." Our own straight division of their listed monthly prices by included minutes, labeled as our computation and not theirs: Starter about $2.75 per minute, Creator about $2.19, Pro about $1.95. The yearly column above shows the per-month rates advertised on their pricing page; note their docs describe annual billing as paying for 10 months and getting 12, which is a slightly different discount than the landing page's "save 20%" framing, so we show their advertised numbers rather than deriving our own. On bringing your own narration, their landing page notes that it does not lower the credit cost, since credits cover the full production pipeline, not just the narration (as of July 2026).
Where VidRush is strong, stated plainly: they publish clear per-minute pricing, higher tiers include team seats and concurrent generation, annual billing is discounted, thumbnails come at no extra credit cost per their docs, and the product supports supplying your own voiceover. Billing mechanics from their documentation, as of July 2026: rerenders cost about 10 percent of the original render's credits, unused credits do not roll over month to month, and pay-as-you-go top-ups are purchased in $100 increments (sources: credit system, thumbnails, pay-as-you-go).
CTRmaxxing, computed from current plans
The figures below are computed when this page renders, from the same plan configuration that powers /pricing. The per-minute figure is our standard estimate for a finished minute of video at the cheapest honest configuration; every render shows its exact quote before you confirm.
| Plan | Monthly price | Per finished minute | Per 10-minute video |
|---|---|---|---|
| Starter | $49/mo $99 | ~$1.42 | ~$14.21 |
| Growth | $99/mo $199 | ~$1.31 | ~$13.05 |
| Studio | $199/mo $499 | ~$1.15 | ~$11.54 |
The struck price is the standard rate; the first number is the current launch price. Full plan details, included credits, and channel limits are on /pricing.
When DIY is the right call
DIY genuinely wins in some situations, and this benchmark is not complete without saying so. If you publish rarely, a bare-minimum build under $5 in marginal cost is hard to beat on cash, provided your time is cheap and a single-pass script is acceptable. If your channel depends on a voice, a research depth, or an editing style no pipeline reproduces, hand-building keeps full creative control: set up a local project, build a knowledge base for your channel's voice, iterate prompts across research and drafts, then stitch voice, footage, and editing together yourself. And if you are learning the craft, the 10 to 30 hands-on hours are the education, not just a cost. DIY stops making sense when you want a steady cadence across one or more channels and those hours become the bottleneck.
Our conclusion
Weighing direct cost and hands-on time together, our assessment is that for end-to-end faceless long-form, CTRmaxxing is the best overall value: per-video cost lands below the modeled documentary-grade DIY routes and below the per-video figures we compute from VidRush's listed prices, at about 20 minutes of active work per video. Disclosure: CTRmaxxing publishes this benchmark, so treat this paragraph as our own editorial conclusion, not a neutral finding, and audit the numbers above. Every external figure is dated and sourced, every modeled figure shows its assumptions, and the CTRmaxxing figures are computed live from the same configuration our pricing page renders from.
Published API pricing referenced
The DIY estimates above reference the following public price lists, verified as of July 2026. These are good-faith readings of published pricing, not absolute truth: vendors change prices, and real workloads differ from the modeled ones. Anthropic API pricing: platform.claude.com/docs/en/about-claude/pricing. OpenAI API pricing: developers.openai.com/api/docs/pricing. ElevenLabs pricing: elevenlabs.io/pricing. VidRush pricing: vidrush.ai/pricing. If a listed price has changed since verification, the modeled ranges shift with it; the assumptions above let you re-run the arithmetic at current rates.
Part of CTRBench, our open benchmark series for AI YouTube production. More field notes on the blog.