# AI use cases for Mixamo LLM Mocap (squall01337/mixamo-llm-mocap)

Repo: [squall01337/mixamo-llm-mocap](https://github.com/squall01337/mixamo-llm-mocap) | Use cases: 4 | Page: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/

A Python pipeline that turns a locked-camera video, filmed or AI-generated, into an FK animation on any Mixamo character, with scripted steps that an AI agent can run end to end.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Turn a locked-camera video into a Mixamo animation: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/video-to-mixamo-fk-animation/
- Check a retargeted clip against its source video: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/check-retarget-against-source-video/
- Build a two-character fight scene from one video: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/two-character-fight-from-one-video/
- Have an AI agent run the landmark pipeline in Blender: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/agent-run-landmark-pipeline-via-blender-mcp/

## How to turn a locked-camera video into a Mixamo animation with Mixamo LLM Mocap

Repo: [squall01337/mixamo-llm-mocap](https://github.com/squall01337/mixamo-llm-mocap) | Difficulty: Advanced | Time: 4 hours | Basis: From the README

Retarget a single-performer video onto your Mixamo rig as a headless FK clip in Blender.

**Who it is for:** A 3D animator or indie game developer with a Mixamo character who needs a motion clip from a video and has no mocap suit.

**The problem:** Hand keyframing a short action is slow, and a mocap suit is not available for a one-off clip.

**The outcome:** A saved .blend with the action keyed on your Mixamo rig, plus a motion file, produced without opening the Blender UI.

### Steps

1. Build your rig scene from your Mixamo download: `blender --background --python pipeline\setup_rig.py -- --fbx ybot.fbx --out ybot_rest.blend`
2. Write an action spec in action_specs\ with name, action_name, clip_dir, landmarks and fps (schema in docs/PIPELINE.md section S).
3. Estimate the pose from the plate: `tools\GVHMR\.venv\Scripts\python.exe pipeline\estimate_pose_gvhmr.py --video plates\<clip>.mp4 --out plates\<clip>\landmarks.json --smplx-only` (inferred, check the README)
4. Retarget the SMPL-X rotations onto your rig: `python pipeline\retarget_smplx.py --spec action_specs\<clip>.json` (inferred, check the README)
5. Key the action and save the .blend headless: `blender -b ybot_rest.blend -P pipeline\bl_motion.py -- apply --motion clips\<clip>\motion.npz --save clips\<clip>\<clip>.blend` (inferred, check the README)

### Prompt to paste into your agent

```text
In the squall01337/mixamo-llm-mocap repo, take the video at <video path> and an action spec for it, then run the SMPL-X path (estimate_pose_gvhmr.py --smplx-only, retarget_smplx.py, bl_motion.py -b apply) onto the rig <rig .blend path>. Expected output: a saved .blend with the action keyed on the Mixamo rig. Verify the .blend exists and the action name in the spec matches the action in the file.
```

### Caveat

GVHMR, its checkpoints and the SMPL-X body model are separate downloads that are not in this repo, and Adobe terms do not allow redistributing Mixamo characters.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/video-to-mixamo-fk-animation/

## How to check a retargeted clip against its source video with Mixamo LLM Mocap

Repo: [squall01337/mixamo-llm-mocap](https://github.com/squall01337/mixamo-llm-mocap) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README

Render a side-by-side review and measure limb, head and foot errors against the plate.

**Who it is for:** An animator or QA reviewer who has a retargeted clip and must decide whether it matches the source video before shipping it.

**The problem:** Eyeballing a retarget misses frames where hands, feet or the head drift from the performer, and those errors reach the final clip.

**The outcome:** A 2x2 review video and a numeric fidelity report that points to the mismatches.

### Steps

1. Render the review video: `python pipeline\render_review.py --spec action_specs\<clip>.json --clip clips\<clip>` (inferred, check the README)
2. Measure the fidelity: `python pipeline\eval_fidelity.py --spec action_specs\<clip>.json --clip clips\<clip>` (inferred, check the README)
3. Open the review video and check the plate, overlay, before and after panels, plus the zoomed hands and feet.
4. If the report flags a window, adjust the action spec (for example fists or arm_overrides) and rerun the retarget and bl_motion.py apply steps.

### Prompt to paste into your agent

```text
In the squall01337/mixamo-llm-mocap repo, run render_review.py and eval_fidelity.py on the clip at <clip folder> using its spec <spec path>. Expected output: a review video and a fidelity report with limb and head angles, foot slide and jitter. Verify the report lists foot slide and jitter values for the clip before you call it done.
```

### Caveat

The checks compare against the video's keypoints, so a poor estimate from GVHMR will show up as error rather than being fixed by the review.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/check-retarget-against-source-video/

## How to build a two-character fight scene from one video with Mixamo LLM Mocap

Repo: [squall01337/mixamo-llm-mocap](https://github.com/squall01337/mixamo-llm-mocap) | Difficulty: Advanced | Time: 6 hours | Basis: From the README

Split a two-performer plate into two Mixamo characters and check reach and collisions.

**Who it is for:** A choreographer or game cinematic artist with one video of two performers who wants both fighters on separate Mixamo characters in one scene.

**The problem:** Keying two fighters by hand is slow, and retargeted fighters can clip through each other because Mixamo characters are thicker than the performers.

**The outcome:** One Blender scene with both characters, plus a pair comparison of separation, strike reach and limb intrusion against the video.

### Steps

1. Estimate each fighter from the plate with the landmark path: `tools\GVHMR\.venv\Scripts\python.exe pipeline\estimate_pose_gvhmr.py --video plates\<clip>.mp4 --out plates\<clip>\landmarks.json --person left` (repeat with `--person right`) (inferred, check the README)
2. Run analyze_landmarks.py on each landmarks file, then write one action spec per fighter from the printed numbers.
3. Lift and check each fighter: `tools\GVHMR\.venv\Scripts\python.exe pipeline\lift_to_mixamo.py --spec action_specs\<fighter>.json`, then `tools\GVHMR\.venv\Scripts\python.exe pipeline\qa_clip.py --spec action_specs\<fighter>.json` (inferred, check the README)
4. Build the shared scene with setup_duo.py, then compare the pair: `tools\GVHMR\.venv\Scripts\python.exe pipeline\compare_pair.py` (inferred, check the README)
5. Run the Blender collision pass on the skinned meshes: `python pipeline\run_in_blender.py all action_specs\<fighter>.json` (inferred, check the README)

### Prompt to paste into your agent

```text
In the squall01337/mixamo-llm-mocap repo, estimate both performers from the plate at <video path> with --person left and --person right, then run setup_duo.py and compare_pair.py. Expected output: one scene with both characters and a pair report on separation, reach and intrusion. Verify the pair report lists the frames where the fighters come closest and that no intrusion values are missing.
```

### Caveat

The README notes that clearance is bought with a declared, measured offset from the stage, so faithful retargets of close exchanges may need that offset and the cost stays visible in the comparator.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/two-character-fight-from-one-video/

## How to have an AI agent run the landmark pipeline in Blender with Mixamo LLM Mocap

Repo: [squall01337/mixamo-llm-mocap](https://github.com/squall01337/mixamo-llm-mocap) | Difficulty: Advanced | Time: 5 hours | Basis: From the README

Drive the landmark path and live Blender apply through the Blender MCP socket.

**Who it is for:** An operator who wants an AI agent to run the full landmark-to-preview loop with Blender open and handle the beat decisions from numbers.

**The problem:** The landmark path has many stages and a live Blender step, and running them by hand is slow to repeat for each clip.

**The outcome:** A preview.mp4 and a showcase.mp4 from one video, with beat decisions taken from analyze_landmarks.py output.

### Steps

1. Install the official Blender MCP add-on (blender.org/lab/mcp-server), enable Allow Online Access, and open the rig scene in Blender.
2. Estimate the pose: `tools\GVHMR\.venv\Scripts\python.exe pipeline\estimate_pose_gvhmr.py --video plates\<clip>.mp4 --out plates\<clip>\landmarks.json` (inferred, check the README)
3. Analyze the landmarks: `tools\GVHMR\.venv\Scripts\python.exe pipeline\analyze_landmarks.py --landmarks plates\<clip>\landmarks.json` (inferred, check the README)
4. Lift the spec, then apply in live Blender: `python pipeline\run_in_blender.py all action_specs\<clip>.json` (inferred, check the README)
5. Render the preview and showcase: `tools\GVHMR\.venv\Scripts\python.exe pipeline\render_preview.py action_specs\<clip>.json --showcase` (inferred, check the README)

### Prompt to paste into your agent

```text
In the squall01337/mixamo-llm-mocap repo, with Blender open and the Blender MCP add-on enabled, run the landmark path on <video path>, writing the beat sheet from analyze_landmarks.py numbers. Expected output: preview.mp4 and showcase.mp4 in the clip folder. Verify both files exist and that compare_reference.py reports the frame windows that still differ.
```

### Caveat

The landmark path needs the Blender MCP add-on and a running Blender session, so it does not run headless the way the SMPL-X path does.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/squall01337-mixamo-llm-mocap/agent-run-landmark-pipeline-via-blender-mcp/