What is this?
This is a Codex Skill that turns a topic or script plus an authorized photo of a person into a full AI presenter video: it writes the script, generates voiceover, creates the character, syncs lips, adds subtitles and keyword animations, edits, renders and runs quality checks.
This is a Codex Skill that turns a topic or script plus an authorized photo of a person into a full AI presenter video: it writes the script, generates voiceover, creates the character, syncs lips, adds subtitles and keyword animations, edits, renders and runs quality checks.Is this a standalone app or a Codex add-on?It is a Codex Skill, not a separate app: it plugs into Codex (or any compatible local Skill agent) and picks whatever video, voice and lip-sync capability is available in that environment instead of being tied to one vendor or model.
It is a Codex Skill, not a separate app: it plugs into Codex (or any compatible local Skill agent) and picks whatever video, voice and lip-sync capability is available in that environment instead of being tied to one vendor or model.How do I get it producing a video?Install the Skill into your Codex skills folder, then either tell Codex your topic or script directly in conversation, or initialize a standard job directory first, fill in job.json with the manual checks and upload permissions, and run the pipeline.
Install the Skill into your Codex skills folder, then either tell Codex your topic or script directly in conversation, or initialize a standard job directory first, fill in job.json with the manual checks and upload permissions, and run the pipeline.What does my environment need before running it?A Codex (or compatible local Skill) agent environment, Python 3.9+, FFmpeg and ffprobe, Bash/jq/awk/sed, and at least one callable video-generation, voice-generation and lip-sync capability in that environment.
A Codex (or compatible local Skill) agent environment, Python 3.9+, FFmpeg and ffprobe, Bash/jq/awk/sed, and at least one callable video-generation, voice-generation and lip-sync capability in that environment.What happens if paid video generation keeps failing?After three consecutive failed paid generation attempts, the Skill stops and summarizes the problem instead of retrying endlessly, and it checks for an existing task ID after an interruption to avoid being charged twice.
After three consecutive failed paid generation attempts, the Skill stops and summarizes the problem instead of retrying endlessly, and it checks for an existing task ID after an interruption to avoid being charged twice.Does it store my uploaded images or API keys?No. The repository keeps no API keys, access tokens, signed download links or user material, and task reports strip credentials, temporary URLs and absolute file paths before being saved.
No. The repository keeps no API keys, access tokens, signed download links or user material, and task reports strip credentials, temporary URLs and absolute file paths before being saved.Why are the reference docs split into three separate files?generation.md, editing.md and qa-recovery.md each cover one phase (input/voice/character, timeline/editing, and quality checks), and Codex only reads the file for the phase it's currently in, which keeps context usage low.
generation.md, editing.md and qa-recovery.md each cover one phase (input/voice/character, timeline/editing, and quality checks), and Codex only reads the file for the phase it's currently in, which keeps context usage low.What does the output video look like by default?Vertical 9:16 at 1080x1920 and 30fps, usually 45-75 seconds for topic-based videos, with a clear opening, 2-4 content beats, a simple closing, and loudness normalized to about -16 LUFS.
Vertical 9:16 at 1080x1920 and 30fps, usually 45-75 seconds for topic-based videos, with a clear opening, 2-4 content beats, a simple closing, and loudness normalized to about -16 LUFS.How does it keep the character's lips in sync through the whole clip?The full voiceover track is treated as the timing reference for the entire video, and character footage, subtitles, cuts and keyword animations are all aligned to that audio to reduce lip drift and mismatched cuts.
The full voiceover track is treated as the timing reference for the entire video, and character footage, subtitles, cuts and keyword animations are all aligned to that audio to reduce lip drift and mismatched cuts.What if I don't have a sample of the presenter's voice?If no authorized voice sample is provided, the Skill falls back to a suitable stock voice instead of cloning one.
If no authorized voice sample is provided, the Skill falls back to a suitable stock voice instead of cloning one.What are similar projects?Pixelle-Video, ComfyUI-MiniMax-H3-LongMedia, storycast, video-ai-talking, short-drama-production, aimoneyhunter, creator-flow, eian-riso-broll
Pixelle-Video, ComfyUI-MiniMax-H3-LongMedia, storycast, video-ai-talking, short-drama-production, aimoneyhunter, creator-flow, eian-riso-brollWhich topics is it in?talking-head, digital-human, lipsync, ai-video, audio-generation, subtitles, voice-cloning, video-editing, codex-skill, content-creation, text-to-speech, ffmpeg
talking-head, digital-human, lipsync, ai-video, audio-generation, subtitles, voice-cloning, video-editing, codex-skill, content-creation, text-to-speech, ffmpeg