Narrated video, from script to picture
AI narration that carries the picture.
Paste the script, outline, or episode notes. Project Director casts the narrator, generates the voiceover, times every scene to the spoken line, and assembles a finished narrated video — no mic booth, no editing timeline.
AI Project Director
Conforms picture to script
Voice-model cast
Studio, designed, or cloned
Sidechain mix
Bed ducks under VO
Captions built in
Readable on silent scroll
Two aspect cuts
Wide feed and vertical
Model Overview
Narration playbook + voice-model pipeline
The hardest part of a narrated video is timing. A voiceover sets the clock: clips land where the sentences land, music ducks under the narrator. Project Director's narration playbook inverts the order. Paste the script or episode notes first; the voiceover is read from that text on the narrator you choose - studio TTS, a Fish Audio or CosyVoice design, or a clone of your own - and only then does the agent plan each scene. Narration is generated per scene, so timing survives revisions and a re-take changes one scene without moving the mix.
Open the narration playbookThe voice is generated from your text
Qwen TTS, Fish Audio, CosyVoice, or a clone of your own narrator - the playbook reads your script on the voice you set and keeps it consistent across every scene.
Every shot conforms to a spoken line
The agent splits the copy into segments and times each scene to its audio, so cuts land on sentence boundaries instead of a fixed clip length.
A mix, not a wall of sound
Narration mux keeps ambient audio and sound design under the voice and ducks them while the narrator speaks - then brings them back underneath the pauses.
One narrator across the whole series
Save the voice once and reuse it on every episode or lesson, so returning listeners recognize the channel by ear.
Three rooms this playbook covers
One pipeline, three very different listeners.
The narration playbook is format-agnostic: it conforms visuals to spoken text, so the same run serves a documentary treatment, a lecture, or a podcast recap.
Documentary
A written treatment becomes archival-style B-roll and slow holds, timed to the narrator's pacing.
Classroom
Lecture notes become a lesson in your own cloned voice, with captions students can follow word-for-word.
Podcast studio
Episode notes become a narrated teaser and a wide recap cut - the show's audio stays the product, video stops being extra work.
Why narrated creators build here
A voice you control, visuals timed to the sentence, a real audio mix, and billing that follows the script.
Narration-first, not a template
The script is the spine. Length and pacing come out of the spoken words, so the video is exactly as long as the argument needs.
Choose or clone the narrator
Design a voice with Fish Audio, clone your own with CosyVoice, or start from a studio TTS voice - then reuse the same narrator across the series.
Sound design that listens
Ambience and effects ride under the voiceover and duck automatically while the narrator talks; go narration-only with replace mode when the picture needs silence around the words.
Re-take the shot, keep the take
Fix the one scene that drifted. Narration timing and the mix elsewhere are untouched, and you pay credits only for the single run.
Works loud and works muted
Captions and on-screen text are produced in the same run, so the lesson lands whether or not the viewer pressed unmute.
Priced like production, not software
No mic booth, no stock library, no per-seat editor. Credits burn when the agent writes, voices, or renders - and nowhere else.
Clean for monetized channels
Outputs export without a watermark and are usable commercially under Moosky and the selected model-provider terms.
How it works
From script to narrated video in four steps
Paste the words, cast the narrator, and let Project Director conform, mix, and export the finished video.
1. Paste the script or notes
Drop in the voiceover, lecture outline, or episode beats. The playbook treats your text as the master document - it narrates what you wrote, not a rewrite of your script.
2. Cast the narrator
Pick a studio voice, attach a sample, or select a clone you saved earlier, and set the tone - measured documentary, warm lesson, punchy teaser.
3. Scenes conform to the voice
The agent generates the narration, splits it into segments, plans a visual per beat, and muxes the track with ambient sound ducked under your words.
4. Review, re-take, export
Watch the cut in chat, re-take the one shot that drifted, then pull a wide upload and a vertical Short from the same timeline.
Credit-based
Simple, honest pricing
A ten-minute narrated video - roughly a thousand five hundred spoken words - commonly lands between $25 and $70 in credits depending on scene count and re-takes; the script pass and the narration quote are cheap, the picture is where the budget goes. No subscription: you pay when the agent reasons, voices, or renders.
More from Moosky AI
Explore our full suite of AI tools.
Frequently Asked Questions
Does this make narrated video or just audio?
Video. The voice models generate the narration track, but the run does not stop there: Project Director plans a visual for every spoken segment, muxes the audio onto the picture with the bed ducked under the narrator, and exports a finished file. If you only want the audio bed, the standalone text-to-speech pages cover that.
Will it rewrite my script before recording it?
No - the narration playbook reads your text. Paste a finished voiceover for word-exact delivery, or paste talking points and the playbook expands them into spoken segments you approve in chat before any audio renders.
Can the narrator sound like me - or the same on every upload?
Both. Upload a short sample to clone your own voice, or design one with tools like Fish Audio or CosyVoice, then save it. The same narrator is reused across every episode and lesson in the series, so the channel keeps one voice.
How long can a narrated video be?
The playbook is built for multi-minute runs - five, ten, fifteen minutes of spoken copy - because each scene is its own generation and quote. Long treatments work best when you batch: lock the narrator and style once, then produce section by section.
What happens to the music and sound effects when the narrator talks?
In mix mode the ambient bed, sound design, and diegetic audio keep playing at full level and duck automatically under the voice while a sentence is spoken, then recover in the gaps - the same sidechain pass an audio engineer would ride by hand.
One scene came out wrong. Do I redo the video?
No. Re-take the single scene; its new footage replaces that shot while the narration, the timing, and the mix everywhere else stay exactly as they were. You spend credits for that one run.
Can course and client work use these videos?
Yes. Exports carry no watermark and are usable commercially under Moosky and the selected model-provider terms - paid courses, client explainers, monetized uploads are all within the license.