Remotion Video Studio

1 - Your video

Drop your video here

MP4, MOV or WebM. Cloud transcription uploads extracted audio; local mode keeps audio on this device.

2 - Transcript

What is spokenauto

Code-switched speech. Nepali is transcribed in Devanagari, English words stay in Latin - the studio loads a Devanagari face so both render.

Speech engineauto - this device

Gemini first, then Groq, NVIDIA, and this device if unavailable. Cloud providers receive only extracted audio. Any fallback is reported.

Speech model465 MBBest multilingual quality this studio can run on-device. Downloaded once, then kept in this browser.

Spoken language

AI cleanup with NVIDIA (optional)Sends transcript text to NVIDIA for spelling and punctuation changes. This may change spoken wording and estimate new timings. Leave off for verbatim captions.Keep English words in EnglishNepali speech is code-switched, and a recogniser told the language is Nepali writes the English it hears in Devanagari anyway - कम्प्युटर for computer, बैंक for bank, ओटिपी for OTP. Those go back into Latin; a Nepali word never does, and a word carrying a Nepali ending is left alone.

Whisper runs inside this tab with WebAssembly. The audio never leaves the machine, and every word gets its own timestamp for karaoke styles.

AI shortcut

AI Caption Templatecopy & use

Already have a transcript, or want a bespoke look the panels below can't produce? Download this template, paste it and your transcript into an AI, and upload the .tsx it returns. It runs the same rendering engine as this studio - word timing, balanced line breaks, mixed Devanagari/Latin fonts, the legibility scrim - so the result looks and behaves exactly like a caption built here.

You are editing the AI Caption Template for a Remotion video project (ai-caption-template.tsx). Generate a complete, ready-to-render .tsx file that burns subtitles into my video.

My video: [a public https:// URL, or staticFile('your-file.mp4') if you will upload the video file alongside this one]
Dimensions: [WIDTH]x[HEIGHT] at [FPS] fps, [DURATION] seconds long

Transcript (paste an .srt or .vtt with timestamps if you have one - this studio's Auto or Write tab can export one - otherwise paste plain text and time it for me):
[PASTE YOUR TRANSCRIPT HERE]

Spoken language(s): [e.g. English / Nepali / Nepali+English code-switched]
Caption style: [e.g. bold social captions with word-by-word yellow highlight, Anton font, bottom third / clean broadcast subtitles on a translucent bar, Inter font / karaoke-style with the current word glowing]

Follow the AI EDITING CONTRACT in the attached file exactly. Reply with ONE complete, runnable .tsx file - no diff, no explanation.

Template

0 captions

fps

Drop in a video to subtitle

During transcription, your video stays local and cloud engines receive only extracted audio. Your edits are saved to this browser as you work.

1UploadMP4, MOV or WebM, or paste a link - then transcribe, paste a script, or import an .srt
2EditA full-size preview with the timeline beside it
3Style & tools18 presets, per-caption sound, objects, bulk transcript edits
4RenderBurn the subtitles in and download the video

Caption look

Font102 of 102

heavy, loud, poster-like. Good for social hooks, big single-word statements, sports and hype.

Scriptno text yet

The Latin face draws English words and the Devanagari face draws Nepali ones, chosen per character - so a mixed line stays in one visual voice.

Typography

Size5.6% of height

Weight (Anton has one)400

Letter spacing0px

Line height1.2

Letter case

Alignment

Lines per caption2

Fill

Text fill

Text

Spoken word

Highlight the word being spoken

Spoken word

Karaoke wipeThe spoken word fills left to right across its own timing instead of switching colour on one frame.

Motion on the spoken word

The spoken word only changes colour or scale.

Always emphasise0 words

Emphasis

These words keep the emphasis colour every time they appear - the offer, the price, the brand name - matched without case or punctuation.

Effects

Outline11%

Outline colour

Drop shadow0.6

Shadow colour

Glow0

3D depth0

Tilt0deg

A hard-edged depth stack instead of a blur: it stays crisp when the video is scaled, which is what makes chrome, comic and retro looks read on a phone.

Backdrop

Legibility scrim0.2

A gradient wash that fades in behind the caption zone - the fix for white type disappearing over sky, snow or a bright background.

Placement & motion

Reveal

Each word pops in on its own timestamp - the social-caption look.

Entrance

Stamp, Whoosh and Glitch also work with the word-by-word reveal - and the Sound tab can match an effect to whichever one you pick.

Distance from edge17%

Block width82%