Rodeo FX has put a number on its season 3 of House of the Dragon. 394 shots, six CG dragons, and a Battle of Tumbleton running to hundreds of shots with roughly a hundred of those carrying crowds, per The Art of VFX’s write-up of the studio’s breakdown, which went up this week alongside a reel. Four days ago the best public record of who solved what on that season was one conference talk about scaling Tumbleton’s fire. Now there is a shot count.
394 from a season worked by eight vendors. That is a large slice for one house, and it is worth knowing what sits inside it before reading it as a league table — set extensions for King’s Landing, fuller CG for the market town itself, and the dragons.
The dragon carrying the finale was built two seasons ago
Silverwing was designed and animated by Rodeo FX for season 2. Season 3 flies her over Tumbleton and sets her on an army.
This is how a six-dragon season gets paid for. Build one hero creature properly once — skeleton, wing membrane, scale layout, shaders that hold from a mile out to a close-up — and the cost lands in one season while the value spreads across several. Six dragons on screen and six dragons in production are different counts, and nobody says so in a press release.
Reuse is not free, though, and anyone who has inherited an asset knows where the bill arrives. New shots mean new camera distances, and a membrane that read fine in a wide from season 2 will show its displacement at half the range. New lighting means the shaders get re-tuned. New choreography means the rig meets poses it was never tested against, and something in the shoulder breaks. The saving is real. It is a saving on the first eighty per cent.
Crowd, fire and collapsing buildings in the same frame
About a hundred crowd shots inside the Tumbleton sequence, layered with dragonfire and large-scale destruction simulations.
Those three systems in one shot are the expensive intersection in all of this, because each one wants to run after the others. The destruction needs the fire’s forces to know what to push. The fire wants the geometry the destruction produced, or it burns through a wall that is no longer standing. The crowd has to react to both — to flinch at the right frame, fall in the right direction, be somewhere else entirely once the roof has gone. And every time an artist changes one, the caches downstream of it go stale.
You can solve that with brute force and a long calendar. Or you solve it with workflow, which is precisely the word Rodeo FX’s FX supervisor chose for the title of his Houdini HIVE talk last month, and it reads differently now that the shot count is public. A hundred crowd shots is a hundred opportunities for a re-sim to invalidate three other departments.
The rider is a real person
The breakdown calls out plate integration for Silverwing’s rider, and it is the sort of line that passes without comment until you think about what it covers.
An actor is filmed on a motion rig against blue, given a performance by a gimbal operator working from previs. The dragon then gets animated — and animation will change, because notes exist. By the time the shot is final, the body is reacting to a flight path that may not be the one it was pitched and rolled to on the day. Closing that gap is warping, re-timing, and in the worst cases re-projecting the actor onto geometry that follows the dragon the animators settled on.
Lighting is the other half. A person lit on a stage and a dragon lit in volumetric sky do not match until someone makes them, and the tell is usually the edge of the hair.
What the record still does not say
Wētā FX has also published on season 3. Two vendors out of eight, which beats the one conference talk and a general featurette this season had a week ago, and is still thin for a production whose other showpiece — the naval battle where the ships were the hard part — has no per-vendor record at all.
What a reel cannot show is the order things were run in. The shot count tells you how much Rodeo FX delivered. The sim dependency graph behind those hundred crowd shots — who cached first, what got frozen to let the others finish, which passes were given up on and comped instead — is the part that would be worth something to the next crew who has to put an army under a dragon.