Runs great
RouteCUDA with the optional NVIDIA speed pack (1.6 GB), or DirectML with no extra download
About s to make 1 minute of narration with the speed pack, about s without (measured on an RTX 5070 Ti). A 10-minute chapter: about to minutes.
Coming soon · Windows first, then Mac
Hindi narration, made on your PC.
आपकी आवाज़, आपके कंप्यूटर पर।
Write your Hindi script, mark the mood, and generate narration in your own voice, right on your computer. Then shape it like a studio: edit the script, fix one line, move clips on the timeline, add music. No cloud, no uploads.
Generated on your PC Script editor and timeline Windows 10 and 11, 4 GB graphics card
Generated on your PC
The voice model runs on your own graphics card. Scripts, voice clips and audio never leave your computer, and it works offline after a one-time model download.
A script editor made for Hindi
Write and edit in Devanagari, with moods and pauses marked right in the text. Change a sentence and only that sentence is made again.
A real timeline
Every sentence lands on the timeline as its own clip. Move clips, set their volume, pick another take, add music on its own track, and export.
Listen
Made by the app. Not a voice actor.
Every sample below was generated by Narration Studio's own engine from a Hindi script, in one voice, with no editing. Each sentence's mood was set in the script.
Samples are being recorded. Check back soon.
Space plays or pauses · ← → move between sentences
0:00/0:00
Point at or tab to a number to see the words it is read as.
Editor and timeline
Generate it. Then edit it.
The script editor and the timeline are one workspace. Mark a mood where the feeling changes and press Generate: each sentence is made on your PC and lands on the timeline as its own clip. Fix a line, try another take or add music without starting over.
Narration Studio · Homecoming
Script
Voices
Mix
Fast mode
Generate
Calmसुबह की पहली किरण खिड़की से अंदर आई। चाय की ख़ुशबू पूरे घर में फैल गई।
long pause
Happyतभी दरवाज़े पर दस्तक हुई! बरसों बाद भैया घर लौटे थे।
Thoughtfulकभी-कभी इंतज़ार लंबा होता है, पर लौटना हमेशा मीठा होता है।
Voice
My voice 6 moods
Delivery
0:000:050:100:150:200:25
Voice
Music
All 6 sentences made
Saved
- Edited a line? Only that sentence is made again.
- Not happy with a take? Ask for another and pick the best of three.
- Add music or your own audio on its own track.
Features
Everything a narrator needs.
Moods in the script
Mark a sentence calm, happy, sad, angry, fearful or thoughtful. Any mood your voice has.
Your own voice
Record or upload a short clip for each mood. The app clones your voice from them, on your PC.
Continuity
Each sentence carries on from the one before, so the voice never jumps between lines.
Automatic retakes
Every sentence is checked for hiss, whine and pitch jumps. A flawed take is made again.
Pauses where you want them
Short and long pause marks, right in the script.
Numbers read naturally
Years, dates and amounts are read as Hindi words, the way people say them.
A real timeline
Clips on tracks. Move them, set each clip's volume, add music.
Change one line, not the chapter
Edited sentences are marked. Make only those again.
Delivery controls
Speed, emotion strength and continuity, plus a Fast mode for drafts.
Export to WAV
The whole timeline or a selection, ready for YouTube, podcasts or audiobooks.
Private by design
Scripts and voice clips stay on your PC. No account, no cloud.
Never lose work
Autosave, a history of versions, and a trash you can restore from.
Your PC
Will it run on my PC?
Narration Studio runs the voice model on your graphics card. Here is what that takes, measured on real hardware.
Not right? Pick yours:
Should run
RouteDirectML, no extra download
About s per minute of narration on our test card through the same route. We are still measuring AMD and Intel cards.
Below the minimum
The voice model needs about 2.5 GB of graphics memory while it works, so we ask for 4 GB.
Not supported at launch
RouteDirectML on built-in graphics
On a fast desktop's built-in graphics, 1 minute of narration took about minutes. We are working on a lighter model.
Coming later
RouteApple silicon
The Mac version follows the Windows release.
Desktop only
Narration Studio is a desktop app for Windows, with Mac to follow.
Let's find out
Open Task Manager → Performance → GPU to see your graphics card and its dedicated memory, then pick it above.
Checked in your browser only. Nothing is sent anywhere.
Time to make one minute of narration
| Route | Seconds per minute of audio (log scale) | Note |
|---|---|---|
| NVIDIA, CUDA speed pack | RTX 5070 Ti | |
| Graphics card, DirectML | RTX 5070 Ti through DirectML | |
| CPU only | Ryzen 9 9900X3D, not supported | |
| Built-in graphics | Ryzen 9 9900X3D's Radeon graphics, not supported |
| Minimum | Recommended | |
|---|---|---|
| Windows | Windows 10 or 11, 64-bit | Windows 11 |
| Graphics | A graphics card with 4 GB of memory and DirectX 12 (NVIDIA, AMD or Intel Arc) | NVIDIA RTX with 8 GB |
| Memory | 8 GB | 16 GB |
| Disk | About 1.5 GB for the app and voice model | Plus 1.6 GB for the optional NVIDIA speed pack. SSD. |
| Internet | Once, to download the voice model | Once, to download the voice model |
| Mac | Coming later, for Apple silicon | |
Provisional: final requirements are published before launch.
Languages
Hindi first. English too. More on the way.
At launch
हिन्दी Hindi
The whole studio is built around Hindi: sentence splitting, numbers, pauses.
English
Write and narrate English scripts too.
On the way
- मराठी Marathi
- বাংলা Bengali
- ગુજરાતી Gujarati
- ਪੰਜਾਬੀ Punjabi
- தமிழ் Tamil
- తెలుగు Telugu
- ಕನ್ನಡ Kannada
- മലയാളം Malayalam
- ଓଡ଼ିଆ Odia
- অসমীয়া Assamese
Voice models
Two voice models. One studio.
IndicF5
At launchBy AI4Bharat, open source (MIT). Clones your voice from short clips, one per mood, and reads in the mood of the clip you chose. Runs on a 4 GB graphics card.
Included
Indic-Speak
Coming later · ProBy Bodhan AI and AI4Bharat. 44 ready-made voices across 22 Indian languages, with speaking styles built in. Needs a graphics card with 8 GB.
Built with Indic-Speak from Bodhan AI / AI4Bharat.
Each model is used under its own licence. Narration Studio downloads the model on first run.
Your voice stays yours.
- Generated on your PC: scripts, clips and audio never leave it.
- Clone only your own voice, or a voice you have permission to use. The app asks you to confirm.
- No account needed. The internet is used only to download the model.
FAQ
Questions, answered.
When can I get Narration Studio?
It is coming soon for Windows, through the Microsoft Store, with a free trial. A Mac version follows. Get launch news and we will email you the day it is out.
Will it run on my PC?
You need Windows 10 or 11 and a graphics card with 4 GB of memory. NVIDIA cards are fastest. Use the checker on this page to see what to expect from yours.
Do I need an NVIDIA card?
No. Narration Studio also runs on AMD and Intel Arc graphics cards through DirectML. NVIDIA cards can use an optional speed pack that makes narration about twice as fast.
Does my voice leave my computer?
No. The voice model runs on your PC. Your scripts, voice clips and the audio you make stay there. The internet is used only to download the model.
Can I use someone else's voice?
Only with their permission. Narration Studio is for your own voice or a voice you have the right to use, and it asks you to confirm that when you add a voice.
Which languages does it speak?
Hindi and English at launch. Marathi, Bengali, Gujarati, Punjabi, Tamil, Telugu, Kannada, Malayalam, Odia and Assamese are on the way.
How do moods work?
You record one short clip for each mood you want: calm, happy, sad and so on. In the script you mark which mood a sentence or paragraph should have, and the app reads it in that mood, in your voice.
How much will it cost?
There will be a free trial on the Microsoft Store. Pricing is announced before launch, and people on the launch list hear it first.
Your story, in your voice.
Narration Studio is coming soon for Windows. Get launch news and we'll email you the day it's ready.