Turn a Topic Into a Publish-Ready 4K Video — Without a Video Team

If you're a podcaster, YouTuber, or Xiaohongshu creator making video solo — and you've eyed Runway or HeyGen but don't want to be locked into a paid editor — video-podcast-maker is the free, open-source tool to look at first.

It runs inside the AI assistant you already use — Claude Code, OpenClaw, OpenCode, or Codex. It just shipped a major update this morning. And it turns out publish-ready 4K video for YouTube and every major Chinese platform, no video team required.

(If you don't use one of those AI assistants yet, start with Claude Code — it's the easiest way in.)

Snapshot

What it actually is

A topic-to-published-video pipeline that runs inside your AI agent and takes creative direction from you in plain language. You say "make a 5-minute video on X for YouTube." The agent walks you through research, script drafting, voice choice, asset selection, and rendering — asking for your creative decisions at each step. The technical work (video composition, audio mixing, subtitle sync, chapter markers, thumbnail generation) happens invisibly.

Output is 4K by default. Voice can be any of 11 providers including a free default (Microsoft Edge TTS), Azure, ElevenLabs, OpenAI, MiniMax with voice cloning, and 6 Chinese-market platforms. Subtitles stay crisp when the video is compressed for upload.

The project's own instructions page has a section explicitly addressed "For the human reading this (not the AI)" — the maintainer wants a human to polish the script draft rather than accept the AI's first pass. That single design choice says most of what you need to know about the project's opinion: the agent handles technical execution; the human still owns the creative judgment.

What you get

• Topic research — the agent searches the web and gathers source material for you
• Script writing — structured narration with chapter timestamps
• Asset tracking — every image, clip, and sound in your video is recorded with its source and license, so you always know exactly what's in each video and where it came from
• 5 optional add-ons for where your assets come from:
  • Your own files (always free, always local)
  • assetSeeker — license-vetted free stock photos, music, sound effects, icons
  • imagenCN — AI-generated stills and thumbnails
  • videogenCN — AI-generated B-roll (the background footage that plays while you narrate) with a price quote before you spend
  • Hyperframes — transparent overlay animations
• Pricing safety net — any paid step (AI voice, AI image, AI B-roll) surfaces a quote you approve before the agent spends a cent. Nothing runs silently.
• 11 voice providers — free default (Microsoft Edge), plus Azure, ElevenLabs, OpenAI, MiniMax, Google Cloud, and 5 Chinese-market platforms. Switch by naming which one you want.
• Live preview — you can tweak colors, fonts, and timing in a browser preview before exporting
• Ready-made visual blocks — comparison cards, timelines, quote highlights, data bars, animated backgrounds, and about 10 more — so your video looks like a designer made it without touching a design tool
• Bilingual output — Chinese and English script templates, voices, subtitle fonts
• Platform-specific tuning — per-platform script structure, thumbnail aspect ratios, hashtag formats, and call-to-action text baked in for Bilibili, YouTube, Xiaohongshu, Douyin, WeChat Channels

Why this matters for the solo content creator

Three things distinguish video-podcast-maker from the commercial video-AI tools a creator would otherwise consider.

First, you never leave your AI agent. You describe what you want in the agent you already use, and it walks you through decisions. No new interface to learn, no separate subscription, no proprietary editor to master.

Second, the cost model is honest. The pricing safety net is the specific feature a non-coder needs: it cannot silently rack up an AI-video-generation bill. Every paid step surfaces a quote you approve before anything runs. Commercial tools optimize for you not looking at the meter; this optimizes for the opposite. That's exactly what a solo creator's finance sanity depends on.

Third, the Chinese-market platform coverage is a real edge. If you publish to Bilibili, Xiaohongshu, Douyin, or WeChat Channels — or you're an English-speaker who wants to reach those audiences — the per-platform script structure, thumbnail aspect ratios, call-to-action text, and hashtag conventions are already baked in. No English-language commercial video tool does this well.

How it compares to Runway, HeyGen, and the rest

Named alternatives across the video-generation-for-creators landscape:

• Runway — commercial. Best-in-class for pure text-to-video generation. Completely different beat: Runway generates visual footage; video-podcast-maker orchestrates the full narrated pipeline (script, AI voice, subtitles, assets, and composition). Complementary, not competitive.
• HeyGen — commercial avatar-based video. Great for talking-head style; video-podcast-maker is for topic-explainer style with narration and on-screen graphics.
• Synthesia — commercial, enterprise avatar video. Aimed at corporate learning and development teams, with pricing to match.
• Descript — commercial. Excellent for turning podcast recordings into video and for speech editing; owns a different creative workflow.
• Opus Clip — commercial. Trims long videos into short clips. Fits after something like video-podcast-maker — you make the long video here, Opus Clip carves out the shorts.
• claude-blog — The same idea, but for writing blog posts instead of making videos. Built for the same solo creator, and it installs the same way — through your AI agent.

Verdict

Install — with a license nuance the README doesn't emphasize.

For the solo creator, indie podcaster, or small-team marketer already running Claude Code / OpenClaw / OpenCode / Codex: install it. The Chinese-market platform coverage alone is worth the setup time if any of Bilibili / Xiaohongshu / Douyin / WeChat Channels is in your distribution mix. Even if you only publish to YouTube, the free voice + pricing safety net produces a defensible baseline video at ~$0 per video.

The license note: the project is CC BY-NC 4.0 — Creative Commons Attribution-NonCommercial. This licenses the code non-commercially. The videos you produce with it are your own creative output and are not restricted by the tool's license. That's the reasonable reading. But if you're building this into a paid online service where users generate videos through your subscription, you're deploying the code commercially and the license does apply — consult a lawyer before productizing. For personal creator use, monetized YouTube channels, and internal work videos, the license is not the blocker some readers will assume.

This post is what we learned from reading the project. Next post: what happened when we actually installed it and made a video — with real costs and timings, running the latest version end-to-end through Claude Code.

For your agent to install this

You can skip this section — it's the install details your AI agent needs. If you're using Claude Code, OpenClaw, OpenCode, or Codex, say "install video-podcast-maker for me" and the agent will read the tables below.

System / OS

• macOS or Linux (tested on macOS; Linux compatible). Windows is not officially supported — use WSL2 if you're on Windows.
• Python 3.8+ for the TTS script and automation
• Node.js 18+ for Remotion video rendering
• FFmpeg 4.0+ for audio/video processing
• ttsCN component skill (required) — install under ~/.claude/skills/ttsCN or set TTSCN_HOME
• Remotion project — scaffolded via npx create-video@latest (base framework the skill composes on top of)

API keys — required for any use

None. Out of the box, TTS_BACKEND=edge uses Microsoft Edge TTS which is free and requires no key. Free stock assets via assetSeeker are also keyless. You can produce a full narrated 4K video at $0 marginal cost with zero API keys signed up.

API keys — optional (per specific use case)

Only sign up for the ones that unlock a feature the user actually wants. Only the active platform's env vars are read at run time.

Higher-quality or non-Edge TTS voices — pick one:

AI thumbnail / image generation (pick one, or use Remotion-composed thumbnails for free):

AI B-roll generation (only if you install the optional videogenCN component skill) — pay-per-generation via the platform provider; you always see the price before it spends anything.