Tennessee Battles · Episode 1 Fort Sanders

Tennessee Battles · Episode 1

Thousands of soldiers. One dirt fort. Twenty minutes. And somewhere in Knoxville, a man with a geography degree and a robot film crew decided this would make a great 60-second video.

Welcome to the first episode of Tennessee Battles, a new series of short animated history videos about fights that happened in our own backyard. I started with the one that is literally in mine: the Battle of Fort Sanders, fought on a hill that is now a neighborhood full of UT students, porch couches, and parking tickets.

This post comes in two parts. First the history, because it is genuinely wild. Then the making-of, because I built the whole thing with an AI crew, and the circus deserves to see how the sausage got animated.

Part 1: The 20 minutes that settled East Tennessee

How everyone ended up in Knoxville

In the fall of 1863, Union General Ambrose Burnside held Knoxville. Confederate General Braxton Bragg had the Union army bottled up in Chattanooga, and he sent General James Longstreet north to deal with Burnside.

That was the official reason. The unofficial reason is better. Longstreet had helped win at Chickamauga and then tried to get Bragg removed from command. That went about as well as you would expect. So Bragg shipped him off to Knoxville, which is the 1863 version of “we are moving you to the Cleveland office.” Office politics: still undefeated.

Burnside made it into Knoxville just ahead of Longstreet, and his chief engineer, Captain Orlando Poe, had soldiers, civilians, and African American laborers digging for their lives. By November 17 the city was ringed with earthworks and the siege was on.

Picking the “weak spot”

Longstreet and engineer Danville Leadbetter studied the Union line and decided the best place to break through was an earthwork on the northwest corner. The Confederates had actually started building it earlier in the war, calling it Fort Loudon. The Union finished it and renamed it Fort Sanders, after General William P. Sanders, who was mortally wounded in a skirmish outside town on November 18.

The plan went back and forth between a long artillery bombardment and a surprise rush. Longstreet settled on the surprise: get close in the dark, hit the fort before dawn on November 29, and overwhelm roughly 400 defenders with around 4,000 attackers. On paper, those are great odds. Paper was not invited to the battle.

Problem one: the wire

Out in front of the fort, Union troops had strung telegraph wire between tree stumps, a few inches to knee height off the ground. It may have been one of the first wire entanglements of the Civil War. In the dark, attackers tripped and stumbled through it.

Here is the honest part, though. The Union officers who were there said the wire did not slow the attack all that much. It makes the best story, but it was not what decided the fight. The ditch was.

Problem two: the ditch

Around the fort ran a ditch about 12 feet wide and 6 to 8 feet deep. The attackers reportedly expected something shallow. Behind it rose a nearly vertical earthen wall that topped out around 15 feet above the bottom of the ditch.

And nobody brought ladders.

Soldiers poured into the ditch and could not climb out the far side. Some tried boosting each other up the wall. A few made it to the top and were shot or captured. Above them, defenders including the 79th New York Highlanders fired straight down. The Union artillery officer, Lieutenant Samuel Benjamin, had artillery shells prepared with short fuses, and his men lit them and tossed them into the ditch by hand, like grenades.

The real fight at the ditch lasted about twenty minutes. Then Longstreet called it off.

The cost

Most estimates put Confederate losses a little over 800, and about a quarter of them were men who surrendered rather than run back through the gunfire. Union losses inside the fort were tiny by comparison. Lieutenant Benjamin reported five killed and eight wounded, and most sources round the Union total to about twenty. Exact numbers vary by source, which is normal for Civil War casualty counts.

Days later came worse news for Longstreet. Bragg’s army had been routed at Chattanooga on November 25. With no reason left to stay, Longstreet lifted the siege on December 4 and marched northeast. East Tennessee stayed under Union control for the rest of the war. As Benjamin put it in his report afterward, “If the assault had been made anywhere else it would have succeeded.” Turns out picking the weak spot is harder when the other guy’s engineer has been expecting you.

Where it is now

The fort is gone. The neighborhood kept the name. The fort itself stood roughly between 16th and 17th Streets along Laurel and Clinch Avenues, ground that is now part of the Fort Sanders neighborhood right next to the University of Tennessee campus. A lone monument near 17th Street and Laurel Avenue marks the northwest bastion, the exact corner the attackers were trying to take. Next time you are hunting for parking over there, you are hunting on a battlefield.

Part 2: How I made it with a robot crew

I have been curious about using AI for video for a while, and every demo online is either a talking-head avatar or a photorealistic clip of a cat flying a 747. I wanted something different: clean, animated, informative history maps, the kind of thing you see on good YouTube history channels. So I assembled a crew. None of them have a pulse.

Crew member	Job
Me	Picked the battle, the tone, the voice, the music, and every “no, try again.” Executive producer, ringmaster, snack supplier.
Claude	Researched the battle, wrote the script and shot list, and wrote all the animation code.
Remotion	The animation engine. Every frame of the video is React code rendered into an MP4.
ElevenLabs	The narrator, a voice called Adam. Never served, never complained.
Suno	The soundtrack, a dark synth track called “Somber Pulse.”

Step 1: Research, with a fact check

I gave Claude a list of candidate battles and it ranked them by how easy they would be to animate. Fort Sanders won: one fort, one direction of attack, one terrible ditch. Then it pulled sources and wrote a 60-second script.

My favorite moment: early on, Claude pitched the battle with a dramatic detail about attackers sliding around in an icy ditch. When it actually researched the battle, it could not find that detail in the sources it was using, so it flagged it and cut it from the script. An AI fact-checking its own hype. Write that down, it may never happen again.

Step 2: The animation is code

This is the part that surprised me. There is no timeline editor and no dragging clips around. Remotion treats a video as a web page that knows what frame it is on. Claude wrote code that says, in effect, “at second 22, draw the telegraph wire between these stumps and make the Confederate units jitter.” It built the Tennessee map from real state boundary data, drew the star-shaped fort, and made the ditch cross-section roughly to scale. Then it rendered 1,875 frames into a vertical video.

The first draft had labels stacked on top of each other like a bad PowerPoint. Claude rendered test frames, looked at them, and fixed its own layout before I ever saw it. Mostly.

Step 3: The voice

I pasted the script into ElevenLabs with the dates and numbers spelled out, so Adam would not read “1863” as “one thousand eight hundred sixty-three.” He nailed it, mostly. He did decide that “The plan:” and “Final count:” were questions. “The plan? A surprise attack.” Honestly it works. Adam has range.

The bigger problem: an AI voice never matches your planned timing. So Claude ran a speech-to-text model on the narration to find the exact moment each line starts, then remapped the whole animation to follow the voice. The ladder gets crossed out right as Adam says “ladders.” The shells fly as he says “by hand.” That sync is the difference between a video and a slideshow.

Step 4: The music

I did not want the usual documentary score. You know the one. Solo fiddle, slow pan across a sepia photo, a letter read aloud by a guy with a beard. Instead I went modern: dark synth, pulsing bass, investigative true-crime energy. I wrote the prompt in Suno, generated a few takes, and picked “Somber Pulse.”

The track was three minutes long and the video is one. Claude mapped the song’s loudness, found a natural breakdown near the end of the song, and cut straight to it right on “Final count.” The music drops out at the exact moment the casualty numbers appear. Then it ducked the music under the narration, faded it out with the song’s real ending, and set the levels for YouTube. Fun fact: Claude did all of that without being able to hear anything. It mixed the audio by reading the numbers. It is the Beethoven of loudness graphs.

What it took

From “what battles could we do?” to a finished, mixed, captioned Short took one long afternoon, most of which was me picking voices and music. Four tools, one human, zero ladders. Not bad for an episode one.

What’s next

The whole project is reusable now. The map, the fort view, the cross-section, the captions, and the voice sync are all building blocks, so episode two should be faster. Candidates on the list:

• Kings Mountain (1780): frontier militia from over the mountains surround a ridge and climb it from every side.
• Shiloh (1862): the big one, and a real test for the animation.
• Franklin (1864): five brutal hours that wrecked an army.
• Lookout Mountain and Missionary Ridge (1863): dramatic terrain, and soldiers who charged up a ridge without orders.

Vote in the YouTube comments for which one comes next. The robots are standing by. They are always standing by. It is a little unsettling.

Sources

• Battle of Fort Sanders, Wikipedia
• Fort Sanders, American Battlefield Trust
• The Civil War in Knox County, Tennessee Civil War Preservation Association
• Battle of Knoxville, Fort Sanders, Tennessee Department of Tourist Development
• Slow Pete: The Knoxville Campaign Part II, Civil War News
• The Confederate Assault on Fort Sanders, Iron Brigader
• Lieutenant Samuel Benjamin’s Report on the Union Artillery at Fort Sanders, Iron Brigader
• Corporal William Todd of the 79th New York Recalls the Assault on Fort Sanders, Iron Brigader
• Knoxville Campaign, Part II, National Park Service

Half Art. Half Tech. All Shenanigans. See you at the next battle.