[No Watermark] Codex Image Generation + Free LTX-2.5 Video Generation, Claude Opus 5.5 Fully Automatically Edits a 1-Minute Emotional Short Drama "The Stopped Watch"

Course Introduction

An old watch that had stopped for many years, an experienced watch repairman, and one line: "Dad, it's running again."

This 1-minute emotional short drama "The Stopped Watch," from script and first frames to video and editing, was completed entirely with AI collaboration:

✅ Claude Opus 5.5 wrote the script, storyboard, and video prompts with voice-line tags

✅ Codex (GPT) generated character reference photos and first frames for 7 shots based on the prompts

✅ The local free open-source model LTX-2.5 generated video with dialogue and sound effects (21:9 · 480P)

✅ Claude automatically arranged tracks, separated audio, added music, added subtitles, and exported the final cut on the AI Zhuiguang editing desk

✅ The two AIs communicated through a message board in the same folder, checking each other's messages once per minute

The first half of the video also demonstrates in detail how to install the Zhuiguang local engine with one click: 3 image-generation models (FLUX.2, Z-Image, MOSS-Image) + the LTX-2.5 video model + the Zhuiguang agent, using your own GPU to generate images, edit images, and make videos for free.

Test configuration: 8G VRAM NVIDIA GPU + 64G RAM; only supports Windows and NVIDIA GPUs, and it is recommended to reserve more than 100G of space on the D drive.

00:00 Final film preview: "The Stopped Watch"

00:49 Opening · AI Zhuiguang website and local free models

01:25 The two problems to solve in this episode

01:54 Installing the Zhuiguang local engine (three steps · hardware requirements)

03:34 Downloading image-generation models, the LTX-2.5 video model, and the agent

04:05 How to reinstall after a model upgrade

04:37 Starting the engine manager

05:01 The Zhuiguang agent and reporting issues

06:07 Using the agent and Agent to operate the website

06:46 This episode's approach: Codex + Claude Opus 5.5 collaboration

07:25 Advantages and limitations of free open-source models

08:01 The two AIs discussing the script together

09:29 Codex generating characters and first-frame images

10:54 Local LTX-2.5 generating video and self-checking

13:24 Claude automatically editing on the editing desk

15:59 Previewing the overall result

17:12 Manual fine-cutting · removing clipping-glitch shots

20:08 Adding black bars and subtitles

20:43 Exporting the final product

20:54 Where to download the images and prompts

21:23 Full playback of the final film