Claude Codes Influencers

Build a talking mascot that turns any script into a vertical video. Free, open source, no credits.

You write the line. The mascot says it with a real voice, lip-syncs every word, reacts to keywords with poses, blinks, bounces to the audio and gets captions. It outputs a 1080x1920 mp4 ready for Reels, TikTok and Shorts.

Fork the repo, add your mascot's art and a voice, and you have your own AI influencer.

Sparky (hand-built sprite pack)	Kev (code-drawn rig)	Anime (same rig, different style)

---

How it works

 your picture ──► Claude finds the face ──► mascots/<name>/  (mascot.py + sprites/)
                                                   │
 script ──► voice (say / ElevenLabs / Voicebox / mic)
                     │
                     ▼
            lip sync (Rhubarb: sounds ──► 9 mouth shapes) + word timings (Whisper)
                     │
                     ▼
            animator: every frame stacks [ears/tail] + [pose] + [eyes] + [mouth] + captions ──► mp4

The full diagram, and an image prompt for making a poster of it, are in docs/HOW-IT-WORKS.md.

A mascot is just a folder: mascots/<name>/mascot.py (settings) + mascots/<name>/sprites/*.png (art).

---

Quick start (10 minutes)

You need Python 3.10+ and ffmpeg.

# 1. get the code
git clone https://github.com/<you>/claude-codes-influencers.git   # your fork
cd claude-codes-influencers

# 2. install
brew install ffmpeg                 # Windows: winget install ffmpeg   Linux: sudo apt install ffmpeg
pip install -r requirements.txt
bash scripts/install_rhubarb.sh     # the lip sync engine (free, ~20 MB)

cp .env.example .env               # paste ANTHROPIC_API_KEY (for `create`) and ELEVENLABS_API_KEY (optional)

# 3. make a video with one of the example mascots
python make.py video sparky --text "Hey hey! I'm Sparky. Comment the word mascot and I'll send it over!"

Your video is in out/sparky/<time>-<first-words>/sparky.mp4.

The examples talk with the free macOS say voice so they work with zero setup. On Windows or Linux, pass a recording with --audio voice.wav, or use ElevenLabs (step 3 below).

Try the others: python make.py video kev --script examples/script.txt and python make.py video anime --script examples/script.txt.

---

Make your own mascot

There are three ways in. Pick one.

Fastest: upload a picture, Claude rigs it (2 minutes)

python make.py create blobby --image ~/Desktop/blobby.png

Any character picture works: a drawing on a plain background, a transparent PNG, a screenshot of your art. create cuts the character out, puts it on a sprite canvas, then Claude looks at it on a pixel grid and finds the mouth, eyes, skin and outline colours. It renders the talking mouths and blinks on your character, checks its own work on a close-up (up to 3 rounds, fixing anything that's off), and writes mascots/blobby/mascot.py with a matching backdrop, caption colours and a voice description. Then:

python make.py video blobby --text "Hey! I was a drawing two minutes ago."

Needs ANTHROPIC_API_KEY in .env (four Claude calls per mascot, usually well under $1). Busy photo background? Remove it first (macOS: right-click the image -> Quick Actions -> Remove Background) or pip install rembg.

Best quality: Claude Code draws it from your references

Open this folder in Claude Code:

python make.py new blobby --refs ~/Desktop/blobby-front.png ~/Desktop/blobby-side.png

Then tell Claude: "draw blobby from its refs". The draw-mascot skill has Claude study the references and write mascots/blobby/draw.py, which paints the whole character in code: 7 poses, 9 hand-quality mouths, 5 eye states and swinging ears or tails. It compares the model sheet to your references and keeps fixing until they match. That's how Sparky, Kev and Anime were made.

By hand: your own art, step by step

Step 1: start from the template

python make.py new blobby --from-image ~/Desktop/blobby.png

This creates mascots/blobby/ with a settings file and copies your picture in as sprites/pose_idle.png. Use a PNG with a transparent background, roughly 1000-1500 px tall.

Step 2: tell it where the mouth is

python make.py check blobby

Open out/blobby/check/grid.png. It's your picture with pixel coordinates on it. Find the centre of the mouth, then edit mascots/blobby/mascot.py:

MOUTH = dict(
    at=(600, 760),    # <- the mouth centre you just read off the grid
    width=120,        # how wide the mouth opens
    cover=dict(color="#F2C6A0", size=(170, 80)),   # paint over the mouth already in your picture (skin colour)
)
EYES = dict(left=(470, 600), right=(730, 600), size=(52, 64), lid="#F2C6A0")   # optional blinking

Run check again and open out/blobby/check/model_sheet.png to see all 9 mouth shapes on your character.

Give it a voice

Pick one in mascots/blobby/mascot.py:

Voice	Setup	Setting
macOS say	none	VOICE = {"provider": "say", "voice": "Samantha", "pitch": 1.2} (say -v '?' lists voices)
ElevenLabs (best)	cp .env.example .env, paste your API key	VOICE = {"provider": "elevenlabs", "voice_id": "..."}
Voicebox (free, local)	install Voicebox, clone or design a voice, export the line as wav	VOICE = {"provider": "file"} then --audio line.wav
Your own mic	record the line	VOICE = {"provider": "file"} then --audio line.wav

Design a brand-new voice with ElevenLabs (it doesn't copy anyone, it builds one from your description):

python make.py design-voice blobby "a small, excited cartoon creature with a bright, bubbly, slightly squeaky voice, fast and upbeat, clear American accent"
# listen to out/blobby/voice-design/candidate-1.mp3 ... candidate-3.mp3, then keep your favourite:
python make.py save-voice blobby <generated_voice_id>
# paste the VOICE line it prints into mascot.py

Make videos

python make.py video blobby --text "Your script here."
python make.py video blobby --script my_script.txt
python make.py video blobby --audio recorded_line.wav --text "the same words, for better lip sync"

Level up the art

One picture gets you talking. For a mascot that really performs, add more layers to sprites/: poses (pose_excited, pose_point, pose_wave_a/pose_wave_b, pose_shrug, pose_think), hand-drawn mouths (mouth_A ... mouth_H, mouth_X), eyes (eyes_open, eyes_half, eyes_closed) and swinging parts like ears or a tail. mascots/_template/README.md lists every file and what to draw. python make.py check tells you what's missing.

---

Settings cheat sheet (mascot.py)

Setting	What it does
NAME	Display name
CANVAS	Size every sprite PNG shares (None = read from pose_idle.png)
FEET, FACE	Where it stands and where the camera punches in, in sprite pixels
SCALE, PLACE	On-screen size, and where FEET lands in the 1080x1920 frame
MOUTH, EYES	Built-in mouths and blinking for one-picture mascots
PARTS	Ears, tails, antennae that sway and flick (dict(sprite=, pivot=, behind=))
VOICE	say, elevenlabs or file
STYLE	Backdrop gradient, glow, caption keyword colour, caption position
KEYWORDS	Words highlighted in the captions
TRIGGERS	"word": (pose, eyes, seconds, punch_in): what the mascot does when it says that word

---

All commands

python make.py list                                   # your mascots
python make.py create NAME --image pic.png            # picture -> talking mascot (Claude rigs it)
python make.py new NAME [--from-image pic.png] [--refs a.png b.png]   # blank mascot / refs for Claude Code
python make.py check NAME                             # validate art, model sheet, coordinate grid
python make.py draw NAME                              # re-run a code-drawn mascot's draw.py
python make.py video NAME --text "..."                # the whole pipeline
python make.py voice NAME --text "..."                # just the voice
python make.py design-voice NAME "description"        # ElevenLabs Voice Design
python make.py save-voice NAME GENERATED_VOICE_ID
python make.py sheet NAME out/NAME/<take> 0.5 2 4     # stills from a take

Using it with Claude Code

Open the folder in Claude Code and just ask: "make a mascot from ~/Desktop/blobby.png and have it say this script" or "draw a mascot from these references". The repo ships a CLAUDE.md and two skills in .claude/skills/ (mascot-video and draw-mascot) that walk Claude through every step, and Claude Code needs no API key because it uses your Claude plan.

The three examples

• Sparky (mascots/sparky/): a fully drawn sprite pack: 7 poses, 9 mouths + smile, 5 eye states, swinging ears. draw.py paints it in code with Skia; the PNGs are committed, so you only need Skia to change the art.
• Kev (mascots/kev/): a code-drawn rig (rigs/doodle_rig.py) baked into sprites by engine/bake.py. Change STYLE in draw.py to wimpy, familyguy, simpsons, pokemon or anime and run python make.py draw kev.
• Anime (mascots/anime/): the same rig in the anime style. It shows how one rig gives you a whole cast.

Credits

• Lip sync: Rhubarb Lip Sync by Daniel S. Wolf (MIT)
• Word timings: Whisper (MIT)
• Fonts: Luckiest Guy (Apache 2.0) and Oswald (OFL), see engine/fonts/LICENSES.md

The example characters are original drawings. Make your own mascot original too: don't upload characters you don't own.

MIT licensed. Built by KevBuildsApps.