Claude Code text to speech: one script, three voices
Claude Code text to speech for a video voiceover works once the Actionway Skill is installed: paste the script, pick a speech model, and an mp3 lands in your project. Claude Code's own voice mode is dictation — it turns what you say into typed input — so it cannot produce that file. To show the difference between the three models in the catalog, we ran the same 57-word script for a sneaker demo through all of them — MiniMax TTS at $0.002 per 40 characters (text), VoxCPM at $0.002 per 40 characters (text), and Seed Audio at $0.01 per second — and put the three results below so you can listen before you choose.
连接你的 AgentThe script we used
你这样问
Make a voiceover for the sneaker demo video from this script: "Meet the everyday white sneaker. A clean leather upper, a gum sole that grips wet pavement, and a fit that works from the morning commute to the late train home. No logos, no noise, just a shoe that goes with everything you already own. Order today and your first pair ships in three days with free returns."
57 words, 307 characters. The same sentence was sent to each model with no voice, speed, or emotion settings, so what you hear is each model's default.
Listen to the three results
MiniMax TTS · 19 seconds of audio · rendered in 10 seconds · $0.016 for 307 characters at $0.002 per 40 characters (text)
VoxCPM · 19 seconds of audio · rendered in 34 seconds · $0.016 for 307 characters at $0.002 per 40 characters (text)
Seed Audio · 29 seconds of audio · rendered in 30 seconds · $0.01 per second, settled on the audio produced, so the amount is known only after the run
| Model | This script | How it is billed | Length of the result | Render time | What it sounds like |
|---|---|---|---|---|---|
| MiniMax TTS | $0.016 for 307 characters at $0.002 per 40 characters (text) | on the script, in whole blocks of 40 characters | 19 s | 10 s | brisk, even, the most "commercial" read of the three |
| VoxCPM | $0.016 for 307 characters at $0.002 per 40 characters (text) | on the script, in whole blocks of 40 characters | 19 s | 34 s | similar pace, a slightly softer voice |
| Seed Audio | known after the run | on the seconds of audio produced, settled after the run | 29 s | 30 s | slower and more deliberate, ten seconds longer for the same words |
Three things stand out. The two script-priced models bill whole blocks of 40 characters, so this 307-character script is 8 blocks — $0.016 for 307 characters at $0.002 per 40 characters (text) each, $0.032 for 614 characters at $0.002 per 40 characters (text) for both — which is cheaper than spending time deciding. Seed Audio is billed on output and came back a third longer, which matters if your clip has a fixed length: a 19-second voiceover fits a 20-second cut, a 29-second one does not. And none of the three needed a voice picked — the defaults are usable, and the call also takes a speed, an emotion, a language hint, and a reference clip if you want a specific voice.
How one of the runs went, end to end
你说
AGENT 回复
Text to speech. Three models in the catalog, costed for this 307-character script:
你选定: MiniMax TTS
- Sends the 307-character script to MiniMax TTS — it starts as a background job
- Waits on the job; this one finished in 10 seconds
- Downloads the 19-second mp3 into an outputs/ folder in your project (that is where every generated file lands) and hands you the file
结果
The MiniMax TTS voiceover, unedited
1 run · MiniMax TTS · 307 characters, billed as 8 blocks of 40 · $0.016 for 307 characters at $0.002 per 40 characters (text) · charged after completion
That is a real Claude Code session, unedited. Claude Code has voice dictation — you speak and it types what you said — but nothing that speaks back or writes an audio file, so a voiceover you can drop under a video needs a speech model, and the Skill installed with the CLI is what put three of them, with prices, in front of the agent. The run went through Actionway's voice generation capability. Picking a model is the approval — the call runs at the shown price with no second question — and if the wallet cannot cover it, the agent stops before anything is charged and hands you a top-up link with the amount prefilled.
Which one to pick
For a product video with a fixed cut, start with MiniMax TTS or VoxCPM: they are priced on the script, they come back in seconds, and the length is predictable. Pick Seed Audio when you want the slower, more deliberate read and the clip can stretch to fit. The agent does not choose for you — it lists all three with prices and stays neutral; say "you pick" and it takes the best-quality tier. The voiceover is usually one step of a longer chain: the same session can transcribe the file into subtitles, lay a music bed under it, and lip-sync a presenter to the finished track — the sales presenter use case is that chain end to end.
Set it up once
npm install --global @actionway/cli@latest
actionway init
mkdir -p "$HOME/.claude/skills"
cp -R "$HOME/.actionway/skill/actionway" "$HOME/.claude/skills/actionway"You need Node.js 20 or newer and an Actionway account; after sign-in the agent tells you your wallet balance and, if it is empty, links you straight to top-up. Prefer a plugin or a Remote MCP server to the command line? Both end at the same catalog and the same wallet — the get started page lists the steps per client side by side.
常见问题
Does Claude Code have text to speech?Not on its own. Claude Code's voice mode is dictation — your speech becomes typed input — and it neither reads replies aloud nor saves audio. For a file you can ship, the Actionway Skill adds MiniMax TTS and VoxCPM, billed per 40 characters of script, and Seed Audio, settled on the seconds of audio produced.
Which text to speech model sounds best from Claude Code?Listen to the three files above — same script, default settings. MiniMax TTS and VoxCPM give a brisk 19-second read for $0.016 for 307 characters at $0.002 per 40 characters (text) each on this script; Seed Audio is slower, came back at 29 seconds, and settles on the audio produced. Try the two script-priced models first — both together cost $0.032 for 614 characters at $0.002 per 40 characters (text).
How do I generate a voiceover from Claude Code?Paste the script and say what it is for. The agent lists the speech models with prices, you pick one, and it runs the job and saves the mp3 into outputs/. The 19-second voiceover in this guide took 10 seconds to render.