Keep technical terms correct
Every translated line is editable before export. When the dub translates a term students know in English, such as recursion depth, change it back in that block and regenerate only that line.
- Indian languages
- 12
- Every speaker's voice
- Cloned
- Starting price
- ₹40/min
CH 02
HINDI DUB
Use the remote to change the dub. Each channel is the same stand-up set.
What you can dub
AI dubbing for e-learning translates lectures, tutorials and course videos into Indian languages in the instructor's own voice, a faster route to e-learning translation than re-recording each course.
-
Recorded lectures
Long-form classes with one instructor or a team of teachers. The instructor's voice carries into every language edition, and you can correct a term in any line before export.
-
Tutorials and walkthroughs
Step-by-step lessons where every term has to match the screen.
-
Course trailers
Short previews that sell a course in each language.
-
Revision and doubt-clearing sessions
Recorded Q&A where students and teachers both speak.
See the complete dubbing workflow
Follow one video through upload, dubbing, transcript review, line edits and export.
What matters when
dubbing e-learning
Learners stop trusting a course the moment a technical term comes out wrong. A dubbed lesson has to keep the words students will search for, in the voice they already know.
Keep the instructor's voice
Voice cloning carries the instructor's vocal traits into each language, so a Hindi or Tamil edition still sounds like the person teaching the course.
Update a lesson without re-recording
When a slide or example changes, edit the lines that changed and regenerate those blocks. The rest of the dubbed lecture stays as it was.
Review every line before export
Keep terms right and the instructor recognisable across every edition of a course.
Line-level editing
Correct technical terms, formulas read aloud and product names in their timestamped block.
Original speaker voice
Clone the instructor's voice so each language edition sounds like the same teacher.
Regenerate individual blocks
Re-voice one block after an edit instead of dubbing the lecture again.
Translation register
Use Formal for lectures and assessments, or Urban colloquial for informal tutorials.
Multi-speaker separation
Separate co-instructors and guest speakers so each keeps their own voice.
Subtitles
Request an SRT file per language for your learning platform's captions.
Audio mix
Balance dubbed narration against the lecture's background audio before export.
Dubbing API
Dub a course library by creating one job per lesson from your content pipeline.
How dubbing works
Every dub goes through the same four steps in the Sarvam workspace.
Data Structures · Lecture 7 Step 1 of 4 · Upload
Drop a video or audio file
MP4, MOV, MP3 or WAV
dsa-lecture-07.mp40%
165.2 MB · 5:16
- Source language
- English
- Number of speakers
- 1
- Voice
- Clone speaker voice
- Translation style
- Auto
- Target languages
- HindiTamilBengali
- 04:12 · Instructor
A stack follows last in, first out.
स्टैक 'लास्ट इन, फ़र्स्ट आउट' के नियम पर चलता है।
- 04:16 · Instructor
The last item you push is the first one you pop.
जो आइटम सबसे आख़िर में push होता है, वही सबसे पहले pop होता है।
- 04:21 · Instructor Regenerate
Keep the recursion depth in mind.
recursion depth का ध्यान रखें।
- Hindi MP4WAVSRT
- Tamil MP4WAVSRT
- Bengali MP4WAVSRT
Build dubbing into
your workflow
Dub a full course from your content pipeline. Create one job per lesson, list every language you teach in and collect video, audio and SRT for each.
- 12 Indian languages from one upload
- Dubbed video, isolated audio and SRT subtitles for every language
- Every speaker keeps their own cloned voice
Python
import os from pathlib import Path from sarvamai import SarvamAI client = SarvamAI( api_subscription_key=os.environ["SARVAM_API_KEY"] ) media = Path("dsa-lecture-07.mp4") job = client.dubbing.create( source_language_code="en-IN", target_language_codes=["hi-IN", "ta-IN", "bn-IN"], export_options=["video", "audio", "srt"], voice_cloning=True, num_speakers=1, job_name="dsa-lecture-07", ) client.dubbing.upload(job.data.upload_url, media) client.dubbing.start(job_id=job.data.job_id)
Security and privacy
Source files and voice data are encrypted in transit. Sarvam does not use uploaded content to train its models. Sarvam maintains ISO 27001:2022 certification and a SOC 2 Type II report.
- Content controls
- Private
- Source files
- Encrypted in transit
- Voice data
- Encrypted in transit
- Model training
- Not used
Frequently asked questions
Can I review technical terms before publishing a dub?
Yes. Every translated line is editable before export. Correct a name, product term or phrase in its timestamped block, then regenerate only that block.
Can I update one line when a lesson changes?
Yes. Open the dub, edit the translated text of the line that changed and regenerate that block. You do not need to record or dub the whole video again.
Can Sarvam dub long lectures and course videos?
The workspace accepts files up to 2 GB. Through the API, the Starter plan allows 2 GB and one hour per file, Pro allows 3 GB and one hour, and Business allows 4 GB and four hours.
Can videos with multiple instructors be handled?
Yes. Set the number of speakers when you create the dub, and Sarvam separates each voice before translating. Round up if you are unsure, because two speakers merged into one slot blend together in the output. Setup fields lock once the dub is created.
Can I export subtitles with the dubbed lesson?
Yes, through the dubbing API. Add srt to export_options and each target language returns a subtitle file timed to the dub, alongside the video or audio you request.
Dub other kinds of content
Try AI dubbing Start with one lecture. Review the terms in Hindi, Tamil or Bengali before you dub the course.
Try AI dubbing
Start with one lecture. Review the terms in Hindi, Tamil or Bengali before you dub the course.