Vozo Video Translation

CLI prerequisite

Requires the official vozo-cli from npm @vozoai/cli (source: github.com/vozoai/cli). Install with npm install -g @vozoai/cli. Do not install from third-party mirrors or untrusted tarballs. Never print tokens.

Read by task

• Sign in, check the active session, log out, inspect points, or check membership: read references/auth.md. Before creating or downloading projects, establish authentication using the bundled login script, following the result and permission rules in that reference. The script checks storage permissions before checking status or opening a login page; do not start browser login separately.
• Create, query, batch-process, add target languages, chain tools, export, or download: read references/project.md.
• Before creating with a tool not used earlier in this conversation: read references/tools.md for flags, required fields, and plan gating.
• Before presenting a create plan: read references/create-plan-checklist.md and show every applicable resolved value, including defaults.
• For common command sequences: read references/scenarios.md.
• When media may be near a limit, or when estimating points: read references/limits.md.

Before running the login script

Before invoking the login script, first send a separate, user-visible chat message in the user's language:

> I'll check whether you're signed in. If not, a Vozo sign-in page will open; please complete sign-in there. If you don't have an account, you can register for free and receive complimentary points to try the service.

Send this message before starting the script, not after it is running. Do not rely on the script's progress output to inform the user; it can open the page before you have relayed that output. Saying only that you will check sign-in status is not enough.

Translation intent

When a request such as “translate this video” or “convert this video” does not clearly identify what the user wants translated, clarify the intended outcome before choosing a tool. If “convert” could mean changing the file format, clarify that first.

Briefly explain these options in the user's language:

• Translate & Dub: Translate the spoken content and dub it in another language.
• Translate Subtitles: Generate translated or bilingual subtitles.
• Visual Translate: Translate on-screen text such as titles, stylized text, labels, and text in presentation slides.

Ask which outcomes the user wants; allow multiple selections. Do not ask again about outcomes the user has already specified.

Choose the tool based on the intended outcome:

• New dubbing, with or without translated subtitles: use Translate & Dub. When both dubbing and subtitles need to change, use its subtitle removal and translated subtitle options rather than creating a separate Translate Subtitles project.
• Translated or bilingual subtitles while keeping the original audio: use Translate Subtitles. Existing dialogue subtitles belong to this workflow even when burned into the video. If the user wants to replace them, explain the option to remove the original subtitles.
• Titles, stylized text, labels, or text in presentation slides: use Visual Translate.

Distinguish dialogue subtitles from other on-screen text by their purpose, not by whether they are embedded in the video. If unclear, ask the user.

Voice model selection

Complete authentication and the relevant media and points checks before explaining voice cloning models and asking the user to choose.

For Translate & Dub, before invoking any question or choice tool that asks about the voice cloning model, first send a separate, user-visible chat message explaining Auto, VoiceREAL, and VoiceNATIVE using the distinctions below. Write that message in the user's current language and keep the model names in English. Internal reasoning, tool arguments, and option descriptions do not replace this chat message. Only after sending it may you present the model choices. Keep the option explanations below as well.

• Auto: automatically selects the best model; use when the user has no preference about voice, accent, or emotion.
• VoiceREAL: preserves the original voice's emotion and accent; best for expressive content such as dramas, vlogs, and entertainment.
• VoiceNATIVE: delivers a more natural target-language accent; best for clarity-focused content such as ads, e-learning, and explainers.

Keep the model names Auto, VoiceREAL, and VoiceNATIVE in English in every language. For English user-facing choices, use these labels:

• Auto (Recommended): Automatically selects the best model; choose this if you have no preference about voice, accent, or emotion.
• VoiceREAL: Preserves the original video's emotion and accent; best for dramas, vlogs, and entertainment.
• VoiceNATIVE: Delivers a more natural target-language accent; best for ads, e-learning, and explainers.

For other languages, translate only the explanatory text into the user's current language and preserve the English model names. If the user has no preference, use Auto and show it in the create plan; do not require a choice. If the user mentions accent, voice feel, naturalness, or emotion, map the request to VoiceREAL or VoiceNATIVE and explain the selection. Read references/project.md for the full rule.

Interface and product naming

For normal project workflows, follow this skill, its references, and vozo-cli ... --help. Do not inspect CLI source code or replace documented CLI commands with inferred internal APIs unless the user explicitly asks to develop, debug, review, or explain the CLI, or published interface documentation is missing or inconsistent.

Use the website product names in user-facing text: Translate & Dub, Translate Subtitles, Visual Translate, and Lip Sync. Keep CLI commands, flags, response fields, and product names unchanged. Vozo is a brand name and must always remain Vozo (capital V).