Best AI UGC ad generator: one product, one talking clip

The finished clip

Step one: the presenter

Everything downstream is decided here, and it is the step people rush. A lip sync model animates what it can see, so the portrait has to give it a mouth: face square to camera, head and shoulders, mouth fully visible with nothing across the chin, and a neutral closed-mouth expression to start from.

The lip sync guide covers what happens when this step is skipped: a source image where the mouth is hidden produces a completed, charged job and a still portrait with audio playing over it. Generating a portrait on purpose is cheaper than discovering that.

Step two: the line

UGC scripts are short and specific, which suits per-character pricing. Ours ran 94 characters and cost $0.006 for 94 characters at $0.002 per 40 characters (text), which is small enough that testing five versions of a hook is not a budget decision. Speech is billed by the script, so you can price a whole batch of variants before generating any of them.

Step three: putting them together

The lip sync call takes the portrait and the audio and returns the clip. This is the step priced by time rather than by unit — $0.14 per second (720p) on the service used here, with three other lip sync services in the catalog at different rates — and it is the one place where the bill can surprise you. The seconds it charges for are the seconds you ask for in the duration field, not the length of the clip that comes back. Leave duration out and the quote is calculated at eight seconds, whatever your audio actually runs to.

That shape is the argument for keeping the three separate rather than looking for one tool that does all of it. A script change re-runs steps two and three. A new presenter re-runs one and three. A different hook, same face, same length, is the cheapest change you can make — and hook variants are most of what UGC testing actually is.

Running a campaign, not a clip

1. Approve one presenter, then stop generating faces

   Pick a portrait, keep the file, and reuse it for every script. A campaign where the presenter changes between ads reads as stock footage rather than as a person.

2. Write the hooks first, all of them

   Speech is priced per character, so a list of ten hooks can be costed before a single one is spoken. Generate the audio for all of them and listen before syncing any.

3. Sync only the hooks that survived

   This is the timed step, so it is the one to be selective about — and the one where an unstated duration quietly rounds every clip up to eight seconds.

4. Keep the script tight

   Short scripts are better ads and shorter output. Both reasons point the same way, which is unusual and worth taking advantage of.

Running it from Claude Code

Claude Code cannot draw a face, speak a line, or animate a mouth; the Skill installed with the CLI adds image generation, voice generation, and character animation to the same session. Cropping the clip to a platform's ratio, trimming the head and tail, and adding captions are local work the agent does for nothing afterwards. Every run is priced before it starts, and an empty wallet stops the job 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.