Work from a scene list, not a runtime
Every generation on Nidhogg is a short clip of roughly 4–12 seconds — and for explainers that constraint is a feature, because good explainer writing already moves in 5–10 second beats. Write your script first, break it at every idea change, and give each beat one line and one visual. A 90-second explainer becomes a list of ten or twelve concrete shots, each of which is a single, focused generation.
For beats that need continuity — a process shown across two angles, a before-and-after — the multi-shot models, Seedance 2 and Kling 3, chain connected shots inside one generation. Everything else works better as independent clips you order in the edit.
Narration and an on-screen presenter
Generate the voiceover from your script with Nidhogg's text-to-speech engines, which offer a range of preset voices — there's no voice cloning, so choose a preset or record your own narration. If the video needs a human anchor, InfiniTalk turns one portrait photo into a talking presenter synced to your audio track, and it works equally well with a TTS voice or your own recording.
When the script inevitably changes after review, you don't have to regenerate the presenter footage: the lipsync redub tools (LatentSync, Sync Lipsync) re-match an existing talking clip to a new audio track, which turns revision rounds from a rebuild into a swap.
Keeping ten scenes looking like one video
Visual drift is the tell of an amateur AI explainer — every scene rendered in a different style. Prevent it mechanically: use the same model for all scene clips, reuse an identical style clause at the end of every prompt ('clean bright studio lighting, minimal background, soft shadows'), and lean on the LOCKED OFF camera preset for the steady, tripod-like framing instructional content wants.
If a recurring host appears in multiple scenes beyond the talking-avatar segments, train a character from 3–20 photos so the same face renders consistently, and keep their wardrobe description identical in every prompt.
Assembly, captions, and honest limits
Nidhogg generates the components; the final cut is assembled in your video editor — clips in script order, narration underneath, trims at the beat changes. Plan for captions there too: Nidhogg doesn't add subtitles or text overlays automatically, and editor-native captions are the right tool anyway, since most social viewers watch explainers muted.
Two other limits worth knowing before you storyboard: it doesn't record your screen (pair generated scenes with a screen recorder for software walkthroughs), and it doesn't export slides — it generates video scenes and still images, not presentation files.