Rachel Digital Human Production
Jingyi-Wu-Richael/rachel-digital-human-productionReviewed by AI
Checked by RepoPicks Radar on
What is Rachel Digital Human Production?
This is a skill for the Codex assistant, not a program with a screen. You copy the folder into Codex's skills directory, set two API keys, and then ask Codex to produce a talking-head video: it checks that your script, portrait photo and voice sample are usable, clones the voice with MiniMax, builds the video from the photo with HeyGen, makes a 15-second preview and waits for your approval before the full 1080p version, keeping track of each job in a file. Both services are paid and the accounts, credits and permissions are yours to arrange.
How easy is it to install Rachel Digital Human Production?
C · install 1 or 2 tools · Windows
- Windows C
- macOS C
- Linux C
Grade C on all three systems: you must already use Codex, copy the folder into its skills directory ("cp -R rachel-digital-human-production ~/.codex/skills/"), export MINIMAX_API_KEY and HEYGEN_API_KEY, and have Python for the two helper scripts (init_job_state.py, preflight_assets.py). The copy command as written is a macOS or Linux shell line; on Windows you copy the folder yourself.
Rachel Digital Human Production on GitHub
- Stars
- 1,085on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +5stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python, Other tools
- Disk space
- not determined
- Accounts or keys
- MiniMax (voice cloning and narration), HeyGen (image-to-video)
- Cost
- Free, API costs extra · The skill itself is MIT and free. README: "The skill expects the user to provide their own accounts, billing, permissions, and API keys" for MiniMax and HeyGen, and voice cloning plus video generation are billed by those services. The repository states no prices.
- License
- MIT (Permissive)
- Last change on GitHub
- Jul 16, 2026
Rachel Digital Human Production risk signals (5)
- Attentioncloning a voice and face needs that persons consent
The workflow takes "inputs/portrait.jpg" and "inputs/voice-source.mp3" and produces a video of that person speaking a script you wrote. The README's own Safety section says "Use this only for authorized voice, portrait, script, and publishing workflows" and that installing it "does not grant access to ... legal clearance". Using someone's face or voice without their permission runs into image and voice rights in many countries and into the rules of both services.
AI review
- Attentiontwo paid accounts required before anything happens
README, Environment Variables: "export MINIMAX_API_KEY" and "export HEYGEN_API_KEY", with the note that the user provides "their own accounts, billing, permissions, and API keys". Nothing runs locally: every voice clone and every video is a paid call to MiniMax and HeyGen, and the skill explicitly asks for a 15-second preview first to limit that spend.
AI review
- Infothe assistant holds your keys while it works
The keys are exported in the shell where Codex runs the steps, and the skill's stated goal is to "Keep API keys, Authorization headers, signed URLs, private portraits, voice samples, and generated client videos out of the skill package". The assistant still reads those values and your input files as it carries out the playbook.
AI review
- Infomost of the work is instructions not code
The repository holds only .gitignore, LICENSE, README.md, SKILL.md and the folders agents/, references/ and scripts/, and the only scripts named are init_job_state.py (creates a state file) and preflight_assets.py (checks the assets). The actual calls to MiniMax and HeyGen are left to the assistant following the written playbook, so what runs depends on how it reads those instructions.
AI review
- Infoone day of work and no updates since
The repository was created on 2026-07-16 and pushed the same day, with no commits in the 113 days since, no releases, no description and no topics, while carrying 1083 stars and 159 forks. Paid APIs change; nothing here has been adjusted since the first day.
AI review
Evidence is shown in its original language.
What Radar could not check about Rachel Digital Human Production
- Whether the described steps still match the current MiniMax and HeyGen APIs, as nothing has changed since July 2026
- What each video costs, since the repository names no prices and both services bill per use
- What MiniMax and HeyGen keep of the portrait, the voice sample and the finished video
- Whether the assistant really stops for your approval before the full 1080p render, which depends on it following the playbook
- The quality of the resulting video and cloned voice