






Ask for a language and Rask handles the rest: transcript, translation, voices and the mix. Dubbing 2.0 follows the meaning of a line and fits it to the time available.

When the person on camera is the reason people watch, the dub should still sound like them. Ask for a cloned voice and tune the accent and the emotion.

Say "add lip-sync" and the mouth movement follows the dubbed speech, including scenes with more than one person on camera.

Titles, captions and slides carry meaning the audio leaves behind. Ask for them too and the translated text comes back inside the video.

Describe the change the way you would to a colleague. Rask fixes those lines, records them again and keeps the version you had before.

Ask for subtitles burned into the picture or as files, and get the dubbed video, the audio and the transcript back in one place.

rask_get_accountrask_get_usagerask_list_languagesrask_create_uploadrask_import_mediarask_create_projectrask_list_projectsrask_get_projectrask_get_download_linksrask_get_transcriptrask_edit_transcriptrask_retranslate_segmentsrask_list_versionsrask_save_versionrask_restore_versionrask_get_voice_setuprask_list_voicesrask_set_speaker_voicerask_create_voice_clonerask_estimate_costrask_start_dubrask_start_lipsyncrask_start_onscreenrask_start_localizationrask_generate_segmentsrask_get_jobrask_cancel_jobrask_api_operationsrask_api_read# pip install rask-sdk
from rask_sdk import Rask
rask = Rask() # reads RASK_API_KEY
# the price first - free, and it changes nothing
estimate = rask.projects.estimate_project(
media_id=media_id, target_language="es-es"
)
# the price you accepted is the ceiling: a higher one fails instead of charging
created = rask.projects.create_project(
media_id=media_id,
target_language="es-es",
expected_cost_minutes=estimate.total_minutes,
)
processed = rask.jobs.wait(created.id)
outputs = rask.exports.list_project_outputs(processed.result.project_id).data
print(next(o.url for o in outputs if o.kind == "dubbed_video"))// npm install @braskai/api
import { Rask } from '@braskai/api';
const rask = new Rask(); // reads RASK_API_KEY
const order = { media_id: mediaId, target_language: 'es-es' };
// the price first - free, and it changes nothing
const { data: estimate } = await rask.projects.estimateProject({ body: order });
// the price you accepted is the ceiling
const { data: created } = await rask.projects.createProject({
body: { ...order, expected_cost_minutes: estimate.total_minutes },
});
const { data: processed } = await rask.jobs.wait(created.id);
const { data: outputs } = await rask.exports.listProjectOutputs(processed.result.project_id);
console.log(outputs.data.find((o) => o.kind === 'dubbed_video')?.url);API=https://api.rask.ai/v3
AUTH="Authorization: Bearer $RASK_API_KEY"
ORDER="{\"media_id\":\"$MEDIA\",\"target_language\":\"es-es\"}"
# the price first - free, and it changes nothing
CAP=$(curl -s -X POST $API/projects/estimate -H "$AUTH" \
-H "Content-Type: application/json" -d "$ORDER" | jq -r .total_minutes)
# the price you accepted is the ceiling
PROJECT=$(curl -s -X POST $API/projects -H "$AUTH" \
-H "Content-Type: application/json" -H "Idempotency-Key: $(uuidgen)" \
-d "$(jq -c --argjson cap "$CAP" '. + {expected_cost_minutes:$cap}' <<<"$ORDER")" | jq -r .project_id)
curl -s $API/projects/$PROJECT/outputs -H "$AUTH" \
| jq -r '.data[] | select(.kind == "dubbed_video") | .url'# the same work, asked for instead of written
claude mcp add --transport http rask https://api.rask.ai/v3/mcp
claude mcp login rask
> Dub this video into Spanish and keep my voice.
Rask - that will use 4 minutes of your balance. Approve?
Done. The Spanish version is saved next to the original.It is an AI video localization MCP server: the connection between Rask and your AI agent. Once it is in place, you can ask the agent for AI video dubbing, translation, voice cloning, lip-sync and subtitles, and it does the work in your Rask account.
Claude Code, Codex CLI, Cursor, VS Code and Claude Desktop. Connecting takes one command or one block of settings.
Remote. It runs on Rask’s servers and your client reaches it over HTTP at https://api.rask.ai/v3/mcp, so there is nothing to install or keep running on your own machine. A local MCP server talks to one client over stdio on one computer; a remote server over Streamable HTTP serves every client and every machine you sign in from.
In Claude Code it is one line: claude mcp add --transport http rask https://api.rask.ai/v3/mcp, then claude mcp login rask. Codex CLI, Cursor, VS Code and Claude Desktop take the same URL in their own settings, and the documentation has the exact block for each.
Only when the assistant runs without you, on a schedule or inside your own pipeline. Otherwise you sign in through your browser once.
It is the same API v3, reached a second way. The MCP server turns its 82 operations into 29 tools an assistant can call from a conversation, on the same account, the same balance and the same projects. When the work has to run every week without you, call the REST API from your own code instead.
There is no charge for the connection itself. The work uses the minutes on your Rask plan, the same balance the web app uses.
All paid plans: Creator, Creator Pro, Business and Enterprise.
Every paid step is quoted first and never costs more than the amount you approve. You can also set a monthly limit, or give the assistant access that cannot spend at all.
Yes. It sends the file to Rask for you. You can also point it at a link on YouTube, Google Drive or Vimeo.
Yes. You confirm that you have the speaker’s permission yourself, because that statement has to come from a person.
Yes. Titles, captions and slides in the frame are translated and come back inside a new video file.
The assistant starts the job, keeps its id and tells you when the files are ready. The work runs on Rask, so you can close the chat and ask for the result later in the day.
135+ languages and regional options, the same catalogue the Rask app uses. Ask the assistant to list them and it reads the current list from your account.
In the Rask app, under Settings then API keys, every connected client and every key you created is listed. Revoke one and that client stops at once; sign in again whenever you want it back.
It makes one outbound HTTPS connection, to api.rask.ai. That is the host to allow if your network restricts where an assistant may connect.
Rask processes your media in a closed environment and never uses it to train general models. What you ask for also passes through the AI assistant you chose, so that provider handles it under its own terms.