The Small-Studio VFX Pipeline Stack

Every VFX vendor now sells an indie tier, and every indie tier comes with a different set of strings. The prices are on the pricing pages. The strings are buried in the FAQs, and a couple of them can wreck a growing shop. This guide is the software map I wish existed when I was doing this math myself: what a one-to-fifteen-person VFX studio should run in 2026 for compositing, tracking, roto, 3D and FX, rendering, review, and the pipeline glue underneath, with real license prices and the fine print spelled out.

I’ve stood this kind of stack up under pressure before. In 2021 I built a cloud VFX studio on AWS for a heavily 3D music video on a three-week turnaround: Nuke and Flame comps, Maya, camera tracking, roto, and a Deadline render farm, all licensed and running for a remote team. The decisions below are the 2026 version of the decisions that project forced, updated for a licensing market that has changed a lot since then. Mostly in your favor, with a few exceptions worth knowing about.

Last updated: July 2026. There are no affiliate links in this article. Nobody below paid to be here. Software pricing moves, so treat every number as verified at the time of writing.

The $100,000 line

Foundry, SideFX, and Autodesk all draw the same line: gross under $100,000 a year and you qualify for their indie licenses. That one number defines a small studio’s software budget more than anything else, so this guide is organized around it. Three bands:

Solo, under the cap. You qualify for everything: Nuke Indie, Houdini Indie, Maya Indie. Your whole professional stack can run just over $1,100 a year.

The 5-seat shop. The awkward zone. You may still be under the revenue cap, but the indie tiers also cap license counts, and those counts differ per vendor. This is where the fine print starts costing real money.

The 15-seat shop. You’ve outgrown every indie tier, on revenue and on seat count. Now the question is which commercial licenses are worth full freight and which roles can run on the genuinely free tools, which in 2026 are better than they’ve ever been.

Compositing: the Nuke decision

Nuke Indie is $499 a year for the full NukeX node graph, including CopyCat machine-learning training. Commercial NukeX is $5,219 a year, so Indie is about 9 percent of the price for the industry-standard comp toolset. Nuke 17 shipped in February 2026 with BigCat, a training node that scales CopyCat-style roto and cleanup models to hundreds or thousands of frames, plus native import and rendering of 3D Gaussian splats in the USD-based 3D system. If you’re a solo artist under the cap and you comp for a living, this is the strongest deal Foundry has ever offered.

Now the restriction almost every roundup misses: Nuke Indie is one license per organization, not one per artist. The day you hire a second compositor, that person cannot run a second Indie seat under your company. The rest of the restrictions matter too: renders cap at 4096 x 3112 (about 12.75 megapixels in any aspect ratio), which rules out large-format and 6K to 8K plate work entirely. Scripts save in an encrypted format that commercial Nuke cannot open, so there is no migration path later; when you outgrow the license, you rebuild your scripts. There is no render farm or FrameServer support, Python automation is limited to calling 10 nodes, and the license phones home every 30 days. Nuke Indie is a great solo tool and a dead end as a studio foundation, and Foundry designed it that way on purpose.

So what does a multi-seat shop actually do? The 2026 answer is DaVinci Resolve Studio, which includes the node-based Fusion page. It’s $295, one time, per seat. There’s no subscription and no revenue or resolution cap, and Blackmagic hasn’t raised the price since 2021. Do the math on a 5-seat comp team: $1,475 once, versus $19,195 every year for five commercial Nuke seats at $3,839 each. Resolve 21 (June 2026) added Magic Mask v2 AI matting and a Render-in-Place external-matte workflow in Studio. And Fusion inside the free version of Resolve covers trials and overflow seats at exactly zero dollars.

The tradeoffs are real. Fusion is not Nuke: deep compositing is weaker, the hiring pool of Fusion artists is smaller, and big-studio clients who send .nk scripts expect Nuke round-tripping. That’s why the realistic 15-seat pattern is a hybrid: one or two commercial Nuke or NukeX hero seats for the shots and clients that demand them, with Fusion seats carrying the volume. For reference, NukeX at $5,219 includes CameraTracker, Kronos, Furnace, Cara VR, and two render licenses; Nuke Studio, if you conform inside Nuke, is $6,379 a year.

One deadline to plan around: Foundry announced in January 2026 that its entire licensing structure goes subscription-only on January 1, 2027. Existing maintenance and perpetual customers get converted to subscription at their next renewal, with no further perpetual updates. Don’t buy anything perpetual from Foundry expecting it to last.

A quick word on After Effects, because someone will ask. It’s $22.99 a month on the annual plan ($263.88 a year prepaid) or $34.49 month to month, and it’s a motion graphics tool that can comp, not a shot-compositing backbone. If your seats already carry Creative Cloud it earns its keep for mograph and quick fixes. I wouldn’t build a VFX pipeline on it.

Tracking and matchmove

SynthEyes is the pick, and it isn’t close on price. Now a Boris FX product, it runs $325 a year, or $655 for a new perpetual license, which makes it one of the last perpetuals standing in this entire stack. Boris FX shipped SynthEyes 2026 in February, so development didn’t stop with the acquisition, and it exports solves to Nuke, Fusion, Blender, and Houdini. The downsides: the UI is dated and it takes time to learn.

The runner-up deserves naming because the big matchmove departments standardize on it: 3DEqualizer4, which shipped release 8.1 in December 2025 and is a superb tracker. It’s also 7,699 euros for a perpetual license, roughly $9,000, with subscriptions running from 65 euros a week up to 1,704 euros a year. SynthEyes solves the same shots for a small shop at about a fifth of the yearly cost. Pick 3DE if a client pipeline mandates it or you hire a tracker who lives in it. Otherwise the money belongs elsewhere. And if the budget is zero: Blender’s built-in camera tracker is the only no-cost matchmove option in 2026, and it’s fine for straightforward shots.

Roto and paint: do you need Silhouette?

Start with Mocha Pro. The plugin version is $325 a year for a single host family (After Effects, Nuke, or Resolve via OFX), $435 for all plugins, or $655 for standalone plus plugins; Mocha Pro 2026.5 shipped in June. Planar-tracked splines, screen inserts, object removal, stabilization: that’s the 80 percent of daily roto and cleanup work, handled inside the comp app you already run.

Silhouette is the dedicated roto and paint tool, and most small shops don’t need it. At $875 a year for the standalone ($545 for the plugin version, $2,195 perpetual), it earns a seat when roto and paint is a daily department: stereo work, heavy paint and dustbusting, or roto artists who have worked in it for years. Silhouette 2026 shipped in May with new AI tracking, so it’s not standing still. But between Mocha’s planar splines, Nuke’s RotoPaint with CopyCat and BigCat mattes, and Magic Mask v2 in Resolve Studio, the machine-learning tools now cover most of what a generalist shop used to buy Silhouette for. Buy it when the roto workload tells you to, not before.

3D and FX: Houdini Indie is the friendliest deal in VFX

Houdini Indie is $299 a year for the complete Houdini FX feature set: pyro, FLIP, vellum, KineFX, Solaris. The same toolset as a commercial rental costs $3,195 a year. The Karma renderer is included, so your renderer bill is zero. And SideFX is the one vendor whose indie tier acknowledges that teams exist: up to 3 Indie licenses per organization, as long as you’re under $100,000 in revenue and under $1 million in funding over the past 24 months. Houdini is also the industry’s default FX tool, which means freelance FX contractors slot straight into your pipeline without retraining.

The traps are milder than Foundry’s but they exist. Indie saves to its own .hiplc file format, which cannot mix with commercial Houdini licenses in one pipeline, and graduating past the cap is a 10x jump per seat: Houdini Core at $1,340 a year node-locked ($1,995 perpetual) or Houdini FX at $3,195 ($4,495 perpetual). Houdini 21, released late 2025, added Karma Hydra 2 support, Gaussian splat rendering in Karma XPU, and texture baking. Budget for the learning curve; it’s the steepest of anything in this article.

Maya Indie needs a warning label. On paper it’s the same kind of deal: $320 a year versus $1,945 for full Maya, all features, Arnold updates included. But the eligibility terms prohibit work-for-hire for organizations grossing over $100,000, and that describes most client work a VFX shop does. That arguably makes it the most restrictive indie tier of the three. Read the terms before you build a service business on it.

And then there’s Blender, which is free, GPL-licensed, and at this point completely normal to see in a professional stack. Blender 5.0 shipped in November 2025 with full ACES pipeline support and improved HDR color management, 4.5 LTS is supported through July 2027, and 5.2 lands this month. For modeling, layout, and generalist 3D alongside Houdini for FX, the price is hard to argue with.

Rendering: the bill is effectively zero now

This is the part of the stack that quietly became free. Karma CPU and XPU ship with every Houdini license, Indie included. Cycles ships with Blender. And since April 2026, Chaos has made V-Ray for Blender free for commercial work. A small shop in 2026 can render everything it produces without a renderer line item on the budget, which would have been fantasy pricing five years ago.

The paid renderers still have their places. Redshift is $289 a year standalone ($49 monthly) and comes bundled with every Cinema 4D subscription ($839 a year), OctaneRender is 19.99 euros a month billed annually, and Arnold runs about $54 a month on the annual plan (and rides along with Maya Indie). Buy one of these because you’ve standardized on C4D, because a specific look or farm workflow demands it, or because a client pipeline does. Without one of those reasons, the free defaults are enough.

Two caveats. Running two free renderers across DCCs means look-matching discipline that a single paid renderer everywhere doesn’t require. And GPU renderers eat VRAM, so the money you save on licenses has a way of migrating into hardware; the machine side of that equation is its own guide.

Production tracking and review

Shot tracking is where small shops burn money on enterprise tools out of habit, so let’s put numbers on it. Autodesk Flow Production Tracking, the tool formerly known as ShotGrid, is $390 per user per year ($50 month to month). A 5-person shop pays $1,950 a year before it has tracked a single shot. ftrack Studio is $25 per user per month billed annually with 250GB of storage, so about $1,500 a year for the same five seats, with a lighter ftrack Review tier at $10.

Against that, Kitsu is free. It’s open source, built specifically for shot and asset tracking, used by over 300 studios, and it integrates with Ayon, Prism, and Blender out of the box. Self-hosting it means someone handles updates, backups, and uptime, and CGWire sells managed cloud hosting from 99 euros a month for up to 10 users if you’d rather not carry that. For a shop this size, Kitsu is the default, and a paid tool needs a specific reason to replace it.

One caveat that overrides all of this: if a client requires Flow Production Tracking, that’s not a debate. Larger studios mandate it for vendors all the time, its API and automation support run deeper than anything else in the category, and $390 a year per affected seat is a cost of that relationship. Budget it and move on.

Review is a separate layer on top of tracking, not a replacement for it. SyncSketch has a free tier and paid plans around $12 per user per month, and it reviews 3D models and playblasts in the browser, which is what VFX clients need. Frame.io Pro at $15 a month for 5 members and 2TB wins when your clients live in Premiere and After Effects, since Creative Cloud includes Frame.io. And cineSync remains the tool for color-critical synced dailies with a remote supervisor: there’s a free Play tier, with Essential at $19 per user per month annual and Pro at $50. For actually moving the frames between artists, that’s a storage and transfer problem we’ve covered separately.

Pipeline glue: Kitsu plus Prism, or the Ayon bet

Pipeline software is the layer that names your files, versions your shots, and keeps artists from saving “final_v2_FINAL.hip” into the wrong folder. The default recommendation for a small shop in 2026 is Kitsu for tracking plus Prism Pipeline for the pipeline itself, because both are free. Prism 2’s core is LGPL-licensed and free for commercial use, with paid Plus and Pro tiers adding integrations like ftrack, Kitsu, ShotGrid, USD, Unreal, and OpenRV, and you can mix free and paid seats in the same studio, so you only pay for the seats that need the paid integrations.

The integrated bet is Ayon, formerly OpenPype. Its server, pipeline, and APIs are open source under Apache 2.0 with a free self-hosted Community edition, and the 2026 pricing update put most premium addons at a flat 3 euros per user per month on the Pro and Studio plans. Ayon is tracking and pipeline in one system instead of two glued together, which is worth real money in maintenance time if it fits how you work. Either way, plan for the true cost: someone on the team has to own pipeline, and that time is a real line item even when the license fee is zero.

For the render farm, Thinkbox Deadline 10 is still the pragmatic pick, with an expiration date attached. It’s been completely license-free since version 10.1.23, with unlimited workers, and it entered maintenance mode on November 7, 2025: security updates and critical fixes only, while AWS steers new customers toward the paid Deadline Cloud. It works fine today and will for a while. Run it, but start thinking about what replaces it.

Two standards decisions round out the glue, and both are cheap. Color: standardize on OpenColorIO 2.5, the version in VFX Reference Platform CY2026, which is feature-complete for ACES 2.0 and ships built-in ACES 2.0 Studio and CG configs. Scene interchange: OpenUSD earned its first formal Core Specification in 2026 (ISO ratification is in progress), and the 26.03 release added a Gaussian splat schema the same year Nuke 17, Houdini 21, and V-Ray 7 all shipped splat support. Adopt USD file by file as jobs call for it. A pure comp shop should not restructure around it.

The stack on one page

Prices verified July 2026. “Solo” assumes you’re under the $100,000 indie revenue cap; the multi-seat column assumes you aren’t, or won’t be for long.

Add it up for the solo artist: Nuke Indie, SynthEyes, and Houdini Indie come to $1,123 a year, and everything else on the list is $0. That’s a complete professional VFX stack, Nuke and Houdini included, for less than a quarter of one commercial Nuke seat.

FAQ

Can two artists in my company share Nuke Indie?

No. The license is one per organization, not one per artist, so a two-compositor shop cannot run two Indie seats. Your second comp seat is either Fusion inside Resolve Studio ($295 once) or a full commercial Nuke license at $3,839 a year. This single clause is why Nuke Indie works as a freelancer tool and fails as a studio plan.

What actually happens when we cross $100,000 in revenue?

Your indie eligibility ends at renewal, and the jumps are steep: Houdini goes from $299 to $1,340 (Core) or $3,195 (FX) per seat, and Nuke Indie’s replacement starts at $3,839. Worse, your existing project files complicate the move: Nuke Indie scripts are encrypted and cannot open in commercial Nuke, and Houdini Indie’s .hiplc files can’t mix with commercial licenses in one pipeline. Budget the graduation cost before it arrives instead of discovering it mid-job.

Is Fusion really good enough to replace Nuke?

For most invisible-VFX work a small shop bills (cleanup, screen inserts, keys, basic CG integration), yes, and Magic Mask v2 in Resolve 21 closed some of the gap further. Where it falls short: deep compositing, the size of the hiring pool, and clients who expect .nk round-trips. That’s why the multi-seat recommendation is Fusion for volume with a commercial Nuke hero seat when the work demands one, not Fusion for everything.

Do we need Flow Production Tracking (ShotGrid)?

Only if a client mandates it, which does happen and isn’t negotiable when it does. Otherwise Kitsu tracks shots and assets for free, and at $390 per user per year, Flow PT is a hard sell for a shop that gets to choose. If you want a middle path with commercial support, ftrack Studio at $25 per user per month is still actively developed and sold.

Is it safe to build a render farm on Deadline 10 in 2026?

Safe today, not forever. Deadline 10 is free with unlimited workers, and maintenance mode (since November 7, 2025) means security fixes keep coming while new features don’t. Farms run on stable, boring software anyway, so this is fine for the next few years. Just know that AWS’s roadmap points at the paid Deadline Cloud, and have a migration plan on paper before you’re forced to write one.

In conclusion

The 2026 story for small VFX shops is genuinely good: rendering is free, pipeline and tracking can be free, and the indie tiers put industry-standard comp and FX tools within reach of anyone under the revenue cap. Two money questions remain: how you handle multi-seat compositing (the Resolve Studio math is hard to beat) and when the paid tracking and review tools are worth it (usually when a client says so).

And that’s the stack. As always, this is what I’d run based on the work I’ve done and the pricing as it stands in July 2026; your client list and your artists’ muscle memory should outvote any guide, including this one. If you’re weighing a specific combination, leave a comment and I’ll give you a straight answer. The rest of our guides live on the Gear Guides page.