Make podcast scripts and voiceover with AI
Skills used in this guide (3)
HyperFrames 音频相关指引。
npx skills add heygen-com/hyperframes -gHyperFrames 视频的非动画创意方向。用于设计规格（frame.md / design.md）处理、调色板、字体排版、旁白、节拍规划、音频反应式视觉、构图模式，以及品牌与风格决策。原子运动模式与场景蓝图见 hyperframes-animation。
npx skills add heygen-com/hyperframes -g影片与视频编辑、内容创作者在制作视频配音或有声书时，当需要克隆音色、生成情感化配音或流式实时语音合成请用此技能。
Download pack → https://github.com/anbeime/skill/tree/main/skills/tts-voice-synthesis
The pain of podcast/talking-head isn't "no voice" — it's "script reads like a paper, voice sounds robotic, nobody finishes". The point of AI: write the script as spoken human language, pick a natural voice, then decide on visuals — don't pile effects upfront.
Who this is for / not for
For:
- Want a podcast / talking-head but fear script + voice
- Have text to turn into audio distribution
- Want local-runnable voice (privacy-sensitive)
Not for:
- Real-host emotion & presence (AI voice still bounded)
- Copyrighted music / real interviews (that's production)
- "One-click full episode" (quality is per-segment polish)
Before you start
- Topic & outline: what this episode covers, segments, points.
- Voice preference: calm / upbeat / neutral, male/female.
- (Optional) visuals: audio-only or talking-head video.
Steps
1. Install the skill
hyperframes-audio is an audio-generation skill. Copy its folder into your client's skills directory (same as any Agent Skill):
| Client | Directory |
|---|---|
| Claude Code | ~/.claude/skills/ |
| Codex | ~/.codex/skills/ |
| Cursor | ~/.cursor/skills/ |
| Trae | ~/.trae/skills/ |
| CodeBuddy | ~/.codebuddy/skills/ |
| Tongyi Lingma | ~/.lingma/skills/ |
2. Script as spoken language
Turn this article into an 8-min podcast script:
- conversational, like talking to a friend, no written connectors
- first 30s hooks (a counterintuitive point), end with one action tip
- mark where to pause, where to shift tone3. Pick a natural voice
Privacy-sensitive → tts-voice-synthesis local; audition 1 clip, then synth the whole episode:
Synth the script with TTS, neutral-calm voice, 0.95x speed,
auto-pause at commas. Synth the first 1 min for audition.4. Talking-head video adds visuals
Video version → hyperframes-creative for tone, then align audio to captions/visuals.
Pitfalls
| Pit | Symptom | Fix |
|---|---|---|
| Paper feel | Reads like article | Rewrite script conversational first |
| Robotic | Sounds fake | Audition natural voice, slightly slower |
| No hook | Swipe in 30s | Counterintuitive point up front |
| Privacy leak | Content to cloud | Sensitive → local TTS |
| Effect pile | Overshadows | Content first, visuals second |
FAQ
No mic, still podcast?
Yes — script + TTS, no microphone needed.
Choose voice?
Yes — audition 1 clip; local TTS more privacy-controllable.
Audio-only or video?
By channel; video → hyperframes-creative for tone then align.
Script length?
8-12 min steadiest; too long drops completion; hook then substance.
Commercial rights?
Confirm voice/music is licensed for commercial; no unauthorized voice clones.
Related guides
Continue across these scenes:
相关源码
- heygen-com/hyperframes · hyperframes-audio source
- heygen-com/hyperframes · hyperframes-creative source
- anbeime/skill · tts-voice-synthesis source
- Agent Skill open format spec
一句话结论
结论：播客先写"说人话"的口语稿、选自然音色，内容优先画面为辅，别一上来堆特效。
References
Related Scenes
Generate illustrations and marketing images with AI
The worst text-to-image is "one-line prompt, pure luck". The point: set style ref + subject + composition + ratio as structured prompt, then iterate small — not throw "draw something nice" and hope.
Generate icons and article illustrations with AI
The pain of illustration isn't "can't draw" — it's "inconsistent style, blurry when scaled, can't edit". The point of AI vector illustration: set style and size first, want vector not bitmap, so it scales sharp and the source stays editable.
Brainstorm and mind-map with AI
The worst brainstorming is "diverge and stop". The point of AI help: first force enough candidates (quantity before quality), then converge on 3 doable ones by "feasibility / impact" — not stop at "many ideas".
Turn the architecture in your head into a diagram with AI
The pain of diagramming isn't "can't draw" — it's that edits are exhausting: change one box and you re-layout the whole thing. Generating descriptive, re-flowable diagrams turns "drawing" into "describing structure"; editing text is 10x faster than editing the picture.