Best lip sync tools for Claude Code users

Attempt 1: a robot with no mouth

The first run used the robot character from our game asset guide and a line of dialogue generated with the speech service. The job completed, the file arrived, and the result was a robot standing still while audio played. The robot has a screen for a face. There is nothing to animate.

Attempt 2: a mouth hidden behind a scarf

The second run used the desert scavenger from the character generator comparison — a human character, so the mouth problem was solved. Except her scarf sits over the chin in that sheet, and the model animated what it could see: eyes, head, hair. Another completed job, another portrait with sound.

Attempt 3: a portrait made for the job

The third run started with a new image: the same character, head and shoulders, facing camera, scarf pulled down below the chin, neutral closed-mouth expression. One more $0.15 per run image. That is the input a lip sync model actually wants, and the clip below is what came back.

The mouth moves with the audio and the head has a small natural drift. It reads as a character delivering a line rather than a photo with a soundtrack — which is the whole point of the third attempt, and the reason the source image is worth generating on purpose.

Decide where the clip is going first

A talking clip is a piece of footage, and footage has a destination. A store page, a Kickstarter update, a social post, or a trailer beat all want a character delivering a line to camera, and that is exactly what these services produce. In-game portraits and cutscenes are a different question, because those have to sit beside the rest of your art at the same resolution and in the same style as everything else on screen.

The cheap way to settle it is to run one short line, put the result next to the screen it will live on, and look. Six seconds is $0.84 for 6 s at $0.14 per second (720p) — a decision worth making before a whole script is written, not after.

What to check before you spend

1. Is the mouth visible and unobstructed?

   No scarf, no mask, no screen face, no hand near the chin. If your character sheet hides the mouth, generate a portrait that does not — one more image is cheaper than one more lip sync run.

2. Is the face square to camera?

   A three-quarter hero angle animates worse than a straight-on portrait. Face to camera, head and shoulders, neutral closed-mouth expression is the shape these models handle best.

3. Have you tested one line first?

   Run a single line and look at the result next to the screen it will live on before you commission a whole script. Six seconds is $0.84 for 6 s at $0.14 per second (720p) — cheap enough to settle the question early.

4. Is the audio already final?

   The clip is rendered against the audio you supply, so a script change means a new run. Generate the line with the speech service, listen to it, then sync.

Running it from Claude Code

Claude Code cannot animate a face on its own; the Skill installed with the CLI is what puts these services in reach, through the character animation capability. Nothing runs until you pick an option, the price is shown first, and if the wallet cannot cover the run the agent stops before anything is charged.

Set it up once

npm install --global @actionway/cli@latest
actionway init
mkdir -p "$HOME/.claude/skills"
cp -R "$HOME/.actionway/skill/actionway" "$HOME/.claude/skills/actionway"

You need Node.js 20 or newer and an Actionway account. Codex users copy the same folder into ~/.codex/skills/actionway instead, and both routes reach the same catalog and the same wallet — the get started page lists the steps per client side by side.