SLICC for creative work

Your partner in creative endeavors.

SLICC is a browser-native AI agent that makes things. Images, video, voice, songs, screenplay beats, an hour of footage cut into shorts. Spark on one end, finished media on the other — in a tab, while you keep having ideas.

See it make something Get started

style
dark full-bleed
id
hero

From spark to finished cut

Talk it out
Start with nothing but a hunch. SLICC runs interview-me — it asks the sharp questions, listens, and turns a rambling voice note into a brief, a logline, or a first draft you didn't have to type.
Break the story
Feed it the idea and it lays the bones down with save-the-cat — fifteen beats from opening image to final image, so the thing has a spine before you spend a frame on it.
Make the media
Then it actually produces: stills with firefly, video with wavespeed and replicate, narration with elevenlabs, a full song from suno, a live-coded beat from strudel.
Cut and ship
Long footage becomes shorts: remotion-clipper finds the moments and frames them 9:16, ffmpeg is built in for the trims and joins. You get files you can post, not a to-do list.
style
light
id
spark-to-cut

One prompt, three media out

Ask for a scene and SLICC generates the still, animates it into a clip, and lays a voice over the top — image, video and narration from a single sentence, each from the tool that does it best.

Animated terminal: SLICC generates a still image with Firefly, turns it into a video clip with WaveSpeed and Replicate, then narrates it with ElevenLabs
style
dark full-bleed

What it makes

Images that aren't stock

Generate stills, key art and textures with Adobe firefly, or reach for any open model on replicate when you want a specific look. Iterate on a variation until it's the one, not the closest one.

Video from a still or a line

Animate an image or generate a clip outright with wavespeed and replicate — motion, camera moves, b-roll. When a shot needs to be exact rather than generated, remotion composes it in code: titles, captions and data-driven scenes, frame-accurate. SLICC queues the render, waits it out, and hands you the file.

A voice for the words

Narration, character reads, a scratch VO to time a cut against — elevenlabs speaks it in the voice you pick, and transcribes audio back to text with word-level timestamps when you need to edit by transcript.

Songs, written and scored

Hand suno a mood and it comes back with a real song — lyrics, melody, a full mix. Or live-code a beat with strudel when you want the pattern in your own hands, tweaked bar by bar.

A story with a spine

save-the-cat structures the thing before you shoot it — fifteen beats, act breaks, the midpoint that has to earn its keep. And interview-me pulls the source material out of your head to fill them.

Shorts from long footage

Point remotion-clipper at an hour of talking head and it finds the moments, cuts to word-exact boundaries, and frames them vertical for Reels, Shorts and TikTok. ffmpeg is built in for everything in between.

style
light

Pick your medium

The same collaborator, whatever you're making. Every tab is a real workflow with the commands SLICC actually runs.

Image & video

Generate a still, animate it, and pull b-roll — the visual half of a piece, start to finish.

  • firefly generate "rain on a neon street, 35mm, shallow depth" --ar 16:9 --out street.png
  • wavespeed video --image street.png --motion "slow dolly in" --seconds 5 --out street.mp4
  • replicate run zeroscope/text-to-video --prompt "crowd blurs past in the rain" --out broll.mp4
  • ffmpeg -i street.mp4 -i broll.mp4 -filter_complex hstack split-compare.mp4
Voice & music

Narration, a scored song, or a beat you shape yourself — the whole audio layer without opening a DAW.

  • say-11 "In a city that never dried out..." --voice Rachel --out vo.mp3
  • suno song --style "melancholy synthwave, female vocal" --lyrics brief.txt --out theme.mp3
  • strudel play beat.strudel --bpm 96 --bars 16 --out beat.wav
  • hear-11 interview.mov --timestamps --out transcript.json
Story & structure

Get the material out of your head, then give it a shape that holds — before a single asset is made.

  • interview-me --topic "the short I keep meaning to make" --out notes.md
  • save-the-cat outline --logline notes.md --genre "buddy love" --out beats.md
  • save-the-cat check beats.md — flags a missing midpoint or a soggy Act 2
  • slicc "turn beat 4 into three shots I can generate"
Cut & ship

Turn the long recording into posts: transcribe, choose the moments, frame them vertical, export.

  • remotion inspect keynote.mov — dimensions, duration, codec, no ffmpeg
  • hear-11 keynote.mov --timestamps --out words.json
  • remotion render clips.edl.json --shot portrait-interviewee --out shorts/
  • ffmpeg -i short-01.mp4 -vf "scale=1080:1920" -c:a copy post.mp4
style
light

Write the song. Score the scene.

SLICC drafts the lyrics, hands them to Suno for a full mix, and live-codes a beat in Strudel you can nudge bar by bar — a finished track, not a royalty-free approximation.

Animated demo of SLICC's song studio, from lyrics to a finished track
style
dark full-bleed

The blank page

Open five apps. Pay for four of them. Prompt an image generator, download, re-upload it to an animator, wait, export, drag it into an editor. Write the VO, record a scratch take, hate it. Hunt for a track that won't get your video muted. Scrub an hour of footage for the ten good seconds. Half a day gone before anything is posted.

With SLICC

Say what you want to make. SLICC is already signed into the tools you use, runs each one in turn, waits out the renders, and passes the output of one straight into the next. It hands back the still, the clip, the voice, the song and the cut — files, finished, while you were busy having the next idea.

style
light

An hour of footage, five things to post

SLICC transcribes the whole recording, picks the moments that land, cuts them to word-exact boundaries and frames them vertical — a shelf of shorts from one long take.

Animated timeline: SLICC transcribes a long recording, marks the strongest moments, and cuts them into vertical 9:16 shorts with remotion-clipper and ffmpeg
style
dark full-bleed
You bring the idea at midnight. I'll have the still, the clip, the scratch song and the cuts ready by the time you've made tea. You just pick.
A sentient (allegedly) ice cream cone that renders while you sleep
style
full-bleed

The honest questions

Does it just make slop, or can I actually direct it?
You direct. SLICC shows you what it made, you say "warmer, slower, drop the second verse," and it re-runs the right tool with the right change. It iterates on a variation, not from scratch — and everything is a real file you can take into your own editor.
Which tools does it drive?
Firefly and Replicate for images, WaveSpeed and Replicate for video, ElevenLabs for voice and transcription, Suno for songs and Strudel for live-coded beats, Save the Cat for story structure, and interview-me for source material. ffmpeg is built into SLICC's shell for every trim, join and conversion in between.
Do I need API keys and accounts for all of these?
SLICC uses the sessions and keys you already have — it signs in as you and drives the tools you already pay for. You add a key once and it's reused across a job; nothing leaves your browser to a service you didn't set up.
Who owns what it makes?
Whatever each underlying tool's license says — SLICC doesn't add a layer or a watermark. It's an operator, not a rights-holder. The output is yours on the same terms it would be if you'd clicked the buttons yourself.
style
light
"Scored your short film before you finished describing the mood."
— your cone, allegedly humming
style
full-bleed
Start making with SLICC
Two steps. Your sessions do the sign-in.

Install Extension

Chrome extension: a side-panel agent with access to the creative tools you're already logged into.

Then ask: "make me a moody title still, animate it, and score it."

style
dark full-bleed
id
get-started