Ready-made "talking video without a camera". Start from a topic — end with a published video, camera never turned on. Each step maps to a card in the base.
What you get
- A finished 1–3 min video (16:9 for YouTube or 9:16 for Shorts/Reels/TikTok) with voice-over, visuals/avatar, music and subtitles.
- A batch of 3–6 short 9:16 cuts from one long video.
- Ready title, thumbnail and tags for upload.
- A repeatable template: the next video is assembled the same way in a couple of hours.
What you need
Tools from the base: ChatGPT/Claude (script), ElevenLabs or Chatterbox/Kokoro (voice), HeyGen/D-ID or ComfyUI + FLUX (visuals/avatar), DaVinci Resolve or OpenCut (editing), Opus Clip (cutting), vidIQ (title/thumbnail/tags), Publer/SMMplanner (posting). Free path needs only a decent PC (a GPU helps for local voice/images). Fast path: ElevenLabs (~$5–22/mo), HeyGen (free tier, then ~$24/mo), Opus Clip (free tier, then ~$9–19/mo).
Step by step
- Script + hook — in ChatGPT/Claude, use the "Shorts hook (3 variants in 10s)" prompt, pick the strongest, expand into a 60–90s script (hook → retention → CTA). OK when the first line grabs in 2–3 seconds.
- Voice — ElevenLabs (paid, natural) or Chatterbox/Kokoro (free, local). Paste the script, pick a voice, export MP3/WAV. OK when it reads evenly with no robotic jumps.
- Visuals / avatar — HeyGen/D-ID (talking avatar from your audio) or ComfyUI + FLUX (8–15 illustration frames). OK when lips match the audio or frames match what is being said.
- Editing — DaVinci Resolve (free) or OpenCut (in-browser). Lay audio + visuals on the timeline, trim pauses, add quiet music (−18…−24 dB), export 1080p. OK when it never sags and voice sits above the music.
- Cut into Shorts — Opus Clip turns the long video into 9:16 clips with auto-captions; keep 3–6 with the best virality score. OK when each clip stands alone.
- Title / thumbnail / tags — vidIQ generates and scores titles/thumbnails; pick the best, collect tags. OK when the title promises value and the thumbnail reads on a phone.
- Publish & schedule — Publer/SMMplanner: queue the video and cuts, enable cross-posting. OK when posts sit in the queue with correct dates.
Free vs fast (paid)
Free: Chatterbox/Kokoro + ComfyUI/FLUX + OpenCut, manual cutting — needs patience and a GPU. Fast: ElevenLabs + HeyGen + Opus Clip (~$15–40/mo total) — mostly buys speed on the "last mile", not different quality.
Common problems
Robotic voice → shorter sentences, more punctuation, lower stability. Avatar drift → clean audio without music, stay within the free length limit. Wrong pronunciation → spell it phonetically. Sagging middle → cut pauses harder, change shot every 3–5s. Low cut reach → open each cut with a hook, add big subtitles.
Time & money
First video 2–4 hours, then 1–2 hours per video by template. Cost: free (local stack) or ~$15–40/mo for the paid combo.