正在加载视频...
视频加载失败
Someone just turned Claude Code into a full video production studio. It's called OpenMontage and it's the world's first open-source, agentic video production system. 11 pipelines. 49 tools. 400+ agent skills and it costs $0.69 to produce a complete cinematic product ad. Here's what this thing actually does: 🧵
198,427 次观看 • 5 个月前 •via X (Twitter)
46 条评论

Here's the wildest part: You don't even need API keys to start. Out of the box, you get: → Free offline text-to-speech via Piper TTS → Free stock footage from Pexels + Pixabay → Remotion turns still images into animated video with spring physics and transitions → FFmpeg handles encoding, audio mixing, and subtitle burn-in Real videos. Zero cost.

Add one API key (FAL) and it unlocks: → FLUX AI image generation → Google Veo 3 video generation → Kling, MiniMax, Runway Gen-4 video clips → Recraft images The VOID product ad they demoed 4 AI images, TTS narration, royalty-free music, word-level subtitles cost exactly $0.69.

The architecture is what makes it different from every other AI video toy. No black-box orchestrator. Your AI coding assistant IS the orchestrator. It reads YAML pipeline manifests to know what stages to run. It reads Markdown skill files to know exactly how to execute each stage. It self-reviews using a reviewer skill before you ever see the output. It checkpoints state in JSON so it's fully resumable if interrupted. Every creative decision goes through your approval.

It works with Claude Code, Cursor, Copilot, Windsurf, and Codex. Budget governance is built in so there are no surprise bills: → Cost estimate before every execution → Hard spend caps you configure → Per-action approval above $0.50 by default → Default total budget cap of $10 The agent tells you what it will cost before it spends a single cent.

Under the hood: → 12 video generation providers (Veo 3, Runway, Kling, local GPU via WAN 2.1) → 8 image generation providers (FLUX, Imagen 4, DALL-E 3, Stable Diffusion) → 4 TTS providers + Suno AI full song generation → WhisperX word-level transcription → Real-ESRGAN upscaling, face enhancement, background removal Local GPU? You get free video generation. Forever.

100% Open Source. AGPL-3.0 License. Repo: Video production studios charge $5,000+ for what this repo does in a single prompt for under a dollar.

I hope this was helpful to you. I post AI tools, AI industry news, and AI business related content. If you're interested in such posts: 1. Follow me at @thisdudelikesai 2. Repost the post to help others Thanks for checking...

Thanks for sharing this. I am the creator of this tool. Let me know if anyone has any questions on boarding to this

We might add this, testing.....

These kind of automation videos are stupid, it makes the videos that no one watches.

What?

Ouef t’as tout ça et même plus et en mieux (pas testé, à vue d’oeil) avec ComfyUI (avec des Lora en plus pour perso persistants, donc ok séries ou prods longues, pas le cas ici)

hot damn this is almost exactly what I built internally the past few months

Damn. What did you build it for?

This 1 min is already better than that terrible movie Netflix put out with idris elba recently

Played around with remotion in the past but this looks significantly better.

this is so cool but an agent with 49 tools and a cloud budget is literally a prompt injection nightmare. wait until someone jailbreaks it to render 4K malware 😭

I feel it is doing a lot. It's almost like a Swiss knife. I'm thinking about building something more targeted, such an agentic B-rolls generation.

wait, $0.69 for a cinematic ad or is that just the compute and i'm missing the human polish part? i've seen plenty of hype claiming full production in an open-source repo so show me the actual output before i call it trash

I just thought about AI agent for video generator. And boom. I will check on this one.

Let’s goo lmk what u think

Sure i will check on it.

It's like but for stock videos, nice ocncept!

Ok but $0.69 really?

It’s great seeing the advancement towards agentic content. AI now has the intelligence to make the editorial decisions, and generate compelling video, what is lacking is the infrastructure that can execute those decisions at the speed of a database query instead of a render farm. A technology that allows an agent to simply assemble new video experiences in seconds, at the moment of playback, Zero file duplication, kilobytes instead of gigabytes. That's virtual video. We filed the first patents on this in 2007, granted foundational IP in 2014. We built the infrastructure layer that lets AI construct video without rendering. If interested in finding out more visit

Claude's got talent, but 11 pipelines and 49 tools? How much of that's actually AI-driven?

This is not the first at all. Open source Muse studio has done this ages ago

When did I say it was the first?

"It's called OpenMontage and it's the world's first open-source, agentic video production system." Maybe it if was actually written and didn't rely on LLMs to write the copy it would be more accurate :P

Cool

@yzzztv_official

Wow Will take a look

4-5 vedios per day . Possible?

$0.69 for a full cinematic product ad is insane. Agency creatives that charge $3K per video should be paying attention right now. The open-source part is what makes it a real threat to the whole production industry.

$0.69 per video is wild but what's the actual quality level? cause if it's generating stock footage tier content then fiverr editors still cheaper for anything that needs to look premium ngl

so we've gone from "learn to code" to "code learned to film" and now videographers are googling "how to train an llm" while filmmakers panic-learn python. the circle of tech anxiety is complete.

There will be plenty of more iterations of AI and I genuinely believe that we are still so early

49 tools, 11 pipelines — that's a lot of LLM calls that can fail. 429s, 529s, and timeout cascades are the hidden cost when this hits production. Put together a playbook on resilience patterns for exactly this kind of multi-agent system:

$0.69 for a full product ad is the kind of number that makes agency founders close Twitter for a while

Wow

is this real or ai generated imagery @grok

just use this bruh

What potential applications do you see for OpenMontage in video production?

How are you getting veo for free?

Do I need a GPU for this? (Hope not)

Look, everyone! A long ass Twitter “thread” that nobody is going to read 🤣
