Developers / MCP
Bring video translation into your AI workflow.
Use seven MCP tools to prepare and follow a translation. You stay in control of the video you share, the quote you approve, and the files you download.
Connect, confirm, then follow the result
Remote MCP endpoint
https://aitrans.video/api/aiCheck your client first
Your client must support remote MCP and the service's OAuth flow, and be registered by the operator. Send its product, version and exact callback URL through the contact page. Dynamic registration is not available; this guide does not promise compatibility with every client.
Authorize in your browser
Use the endpoint below and your client's normal OAuth connection flow. Sign in and review the requested access on the website. A connection does not share your whole task history or approve spending. Never paste website cookies, access tokens or refresh tokens into a chat.
Share a video and approve its quote
Read the current options and use a returned voice ID. Request an upload link, then choose and share the video on the website. Once the verified upload is ready, prepare a translation and open its original confirmation URL. Creating the task requires website approval of that same quote; changed settings need a new quote.
Keep the original IDs
Create only the confirmed intent, then query that same intent or task. Follow poll_after_seconds and use bounded backoff. UNKNOWN or a lost response does not authorize a second task or another paid submission. A denied or revoked grant needs attention, not repeated creation.
The current MCP creation scope
The example uses English. Read current target_languages, then select target_language and a voice returned for selected_target_language. UI language and detected source language stay separate. Availability does not approve spending; other website settings are not automatically available through MCP.
- Material
- verified_media
- Source language
- auto
- Target language
- Choose from current target_languages
- Mode
- express
- Requested output
- editor
Seven tools, one task
Expand a tool to see its arguments. IDs and request keys below are fictional: use the values returned for your own authorized request. Replace the example voice ID with one returned by get_translation_options.
get_translation_optionsRead supported settings and usable official preset voices.
Example arguments
{}request_translation_uploadGet the website link where you choose and share a video.
Example arguments
{
"request_key": "example-upload-001"
}get_translation_uploadRead the original upload request until verified media is ready.
Example arguments
{
"request_id": "11111111-1111-4111-8111-111111111111"
}prepare_translationGet a server quote and its website confirmation link. This does not approve spending or create a task.
Example arguments
{
"request_key": "example-prepare-001",
"request": {
"source": {
"type": "verified_media",
"value": "22222222-2222-4222-8222-222222222222"
},
"source_language_mode": "auto",
"target_language": "en",
"service_mode": "express",
"preset_voice_id": "example-use-returned-voice-id",
"translation_preferences": null,
"requested_outputs": [
"editor"
]
}
}create_translationSubmit the original, website-confirmed intent.
Example arguments
{
"intent_id": "33333333-3333-4333-8333-333333333333"
}get_translationRead the original intent, request key or task using exactly one selector.
Example arguments
{
"translation_id": "job_example_001"
}get_translation_resultRead project and eligible download links, with bounded result metadata.
Example arguments
{
"translation_id": "job_example_001",
"language": "en"
}Read results without guessing
A created task is not completed media. Open the project link to review progress and take any required website actions. Draft generation, unlocking, editing and copying remain explicit website actions; no tool automatically buys or generates a draft.
Downloads open in a browser and require website login. Current ownership, funding, entitlement and output-version checks still apply. A returned link is not a download credential or a promise that a file is available.
result_generation is the eligible task's edit version, or null when it cannot be verified. It is not a content hash or proof of a same-version regeneration. Format and role describe the declared output; an unknown or conflicting actual language stays null. Default calls return no subtitle text or audio. Optional bounded published target subtitles require separate translations:subtitles:read consent and the original grant's task access; see the text guide.
draft_outputs reports at most two persisted draft facts. Its freshness is always unverified: status is not proof of a file or download permission. Follow next_action on the website, especially review_on_website for needs_review.
Example result metadata
{
"result_generation": 0,
"download": {
"key": "publish.dubbed_video",
"format": "mp4",
"role": "dubbed_video",
"language": "en"
}
}Example draft fact
{
"draft_format": "jianying",
"status": "needs_review",
"warning": "compatibility_needs_review",
"freshness": "unverified",
"next_action": "review_on_website"
}What has been checked
One ChatGPT workflow completed all seven tools with a self-authored clip of about 13 seconds, auto source detection (detected as English), English target, Express and a preset voice. Reused media and timing were rebuilt and the unchanged extracted ZIP was accepted by the user in Windows Jianying 11.6.0.14552 on 15 October 2026. This is a specific short-sample result, not acceptance of all languages, modes, voices, clients or desktop versions.
For exact OAuth metadata, recovery rules and client setup details, use the plain-text guide. If authorization refresh is refused, reconnect through your client's browser flow; do not replace credentials manually.
Read the plain-text guide