publikclip

publikclip

open source · on publik

On your phone?

You install publikclip from a computer. Send yourself the link and open it there.

Before you start: A Mac (Apple silicon) or Windows 10/11 x64 PC with ~10 GB free (the speech/audio models download on first run). Scoring runs on publik API by default — your own Gemini API key or a local Ollama model work too

Available for macOS and Windows.

Built on Publik APIBuild yours →

Runs on publik API

AI chat & tools: Chat completions with tool calls, JSON schemas and images, priced per tier in dollars.

No key to paste.
Linking your publik account gives five cents of free use, once.
Then publikclip pays publik's published price per use, and your own key still works.

How pricing works

Vote on publikclip

Read the install guide Open in GitHub

Compared with

OpusClip

Pro plan

$29/mo

What a month of AI costs

OpusClip$29.00/mo
publik API$4.33/mo

publik API: $24.67 a month less than OpusClip.

100 uses a week at the Balanced level · publik API is cheaper up to about 669 uses a week

Make this yoursFork publikclip, change it, publish your version. About 30 minutes. No experience needed.

124 downloads through publik

Having trouble? Tell us

Where publikclip’s AI runs, and what it costs

Prices are for one typical use: one request of about 1,500 words sent and 375 words back. A higher quality score is better.

On your computer

Price: Free per use. Your computer does the work.
Runs with: ONNX Runtime, faster-whisper, Ollama
Memory: Llama 3.1 8B is a 4.9 GB download at 4-bit. gpt-oss 20B runs on systems with as little as 16GB memory. Faster with Apple silicon or a graphics card.
Quality: Llama 3.1 8B: 7

publik API

Price: $0.0100 per use, $10.00 per 1,000 usespublik-balanced, the default level
Quality: MiMo-V2.6-Pro: 46
Setup: Built in. No key to paste; linking your publik account gives $0.05 of free use, once.

Your own key

Price: $0.0013 per use, $1.31 per 1,000 usesMiMo-V2.6-Pro at OpenRouter’s list price, before its fee for buying usage
Quality: MiMo-V2.6-Pro: 46
Setup: Open a provider account, add a card, paste the key into publikclip.

Quality and price, side by side

Quality score and cost per 1,000 typical uses for local models and the three publik API levels
Model	Quality	Per 1,000 uses
On your computer (Ollama, 4-bit download size)
Granite 4.2 3B2.2 GB	9	$0
Phi-4 Mini2.5 GB	6	$0
Llama 3.1 8B4.9 GB	7	$0
gpt-oss 20B14 GB	9	$0
Gemma 4 31B20 GB	19*	$0
Qwen3.5 35B-A3B24 GB	19*	$0
publik API
publik-fastGLM-5.3 Flash	42	$1.00
publik-balancedMiMo-V2.6-Pro · publikclip	46	$10.00
publik-smartGPT-6 Sol	48	$18.00

Quality: Artificial Analysis Intelligence Index v4.3.2, read 2026-09-22 (publik API models 2026-09-25); * = estimated by Artificial Analysis. Sizes: the Ollama library, read 2026-09-22.

What you pay for

On your computer: nothing per use. You pay in disk space, memory and electricity, at lower quality.
publik API: publik’s published price for each use, in dollars, from your publik balance. It is above the model’s cost; the difference runs publik and pays the app’s builder.
Your own key: the provider’s price, billed to an account you keep with the provider.

Install publikclip: publik API is built in How publik API pricing works

How to install publikclip

Every step written out. No terminal experience needed. Pick your setup.

README

Open in GitHub

Long video in. Scored vertical clips out. Everything runs on your machine.

publikclip is an open-source (AGPL-3.0) desktop app that takes a YouTube URL or a horizontal video file and produces vertical 9:16 clips with:

Smart camera — active-speaker-tracked crop paths, smoothed motion, hard cuts on speaker change, punch-ins fired by actual laughter and vocal energy
Word-accurate captions — multiple styles, karaoke highlighting, prosodic emphasis (loud words get loud styling), [laughs] tags from real laughter detection
A virality score you can audit — never a bare number: every clip ships with its subscores, which detectors fired, and every adjustment applied. LLM humor scores get discounted when no actual laughter corroborates them.
Music & SFX — a local, CC-licensed library (import your own or pull from Freesound/Jamendo), auto-suggested from a genre/mood/energy brief derived from what's being said and how it sounds, with manual placement always available and music ducked under speech automatically
Optional real-outcomes loop — connect your own Instagram (via your own Meta app, no middleman) and the scorer calibrates against how your clips actually perform

Every model — speech recognition, forced alignment, diarization, laughter detection, audio tagging, face detection, active-speaker detection — runs locally. The only network calls are the video download and 2–3 small LLM calls (publik API by default — no key to paste; or bring your own Gemini key, or run fully local via Ollama at reduced scoring quality).

Status

Working end to end: hour-long podcast in, rendered/captioned/scored 9:16 clips out, validated on real footage. The Instagram feedback loop ships in-app (sync, clip↔Reel matching, snapshot history, automatic score calibration). Builds are currently unsigned — install from source below, or follow the guided install at publikhq.com/publikclip.

Runs on macOS (Apple silicon) and Windows 10/11 x64. The Windows path is validated on every push by the windows workflow: env resolve, full test suite, NSIS build, silent install, and a launch of the installed app on a clean VM.

Music & SFX

Every clip gets its own music/sfx track, backed by a local library at ~/.publikclip/audio/ (files copied in, plus an index of tags/bpm/licence). The library starts empty — fill it from your own files, or pull Creative-Commons tracks from Freesound/Jamendo straight from the app.

Local import — Library tab → files/folder opens a native picker. Kind (music vs. sfx) and tags are derived automatically; bpm is detected for music.
Online CC sources — the Online tab searches Freesound (sound effects) and Jamendo (music), CC0-only by default. Both need a free API key, saved once in the app: freesound.org/apiv2/apply and devportal.jamendo.com — the panel shows the signup link inline when a key is missing.
Suggestions — suggest audio scores your library against the clip's music brief (genre/mood/bpm/energy) and the event timeline (cut boundaries, laughs), and proposes one music bed plus a few sfx stings. Suggested items show dashed on the track; accept one with its ✓, or delete it — nothing is ever auto-accepted.
Manual placement — drag any library item onto the audio track, drag its edges to trim, drag its body to move it. Select an item for gain, fade in/out, loop, and duck controls.
Ducking — music tagged to duck sidechains under speech automatically at render time (light/medium/heavy, from the clip's brief) so dialogue always stays legible; sfx never ducks.
Licence policy — CC0 by default; a CC-BY item requires opting in ("allow attribution" in the Online tab) and is never silent about it — a credits.txt is written next to every rendered clip that uses one, listing its attribution. NC and ND licences are never fetched, full stop.

Layout

pipeline/   Python package — the entire processing pipeline + CLI
app/        Tauri v2 desktop shell (React UI, Python sidecar)

Development

# pipeline
cd pipeline && uv sync --group dev --group pipeline && uv run pytest
uv run publikclip run "https://www.youtube.com/watch?v=..."

# pick the brain per run: --llm publik (default) | gemini | ollama
# publik API from a terminal, no GUI: PUBLIK_API_KEY=pk_live_... (optional
# PUBLIK_API_BASE_URL, default https://publikhq.com/api/v1); own key:
# PUBLIKCLIP_GEMINI_API_KEY=AIza...

# app
cd app && npm install && npm run tauri dev

License

AGPL-3.0-or-later. Portions adapted from other open-source projects — see VENDORED-LICENSES.md for the full provenance list.