Models
The models Meevra items are written for, where they run and how to prompt them.
Google's recommended default for video generation. Launched at I/O 2026 in the Gemini app, Google Flow and YouTube Shorts; Gemini API GA 2026-08-27 (Interactions API). Text, image or reference to video, conversational editing and extension; audio in every clip.
GA on Vertex AI since 2025-11-17 (veo-3.1-generate-001); the Gemini API still lists preview codes. Use for scene extension, first and last frame control and up to 3 reference images; 4, 6 or 8 s clips with native audio.
Kuaishou
3-15 s clips with native audio and dialogue (Chinese, English, Japanese, Korean, Spanish), Multi-Shot and Custom Multi-Shot, and bound Elements for consistent characters.
Upgraded form of VIDEO O1 for reference-based generation and editing.
Runway
Text to Video and Image to Video, 2-10 s at 720p, 24 or 25 fps.
In-context video editing of clips up to 30 s at 1080p, including multi-shot sequences; release month not confirmed by Runway's page.
ByteDance
30 s clips in one run, up to 30 images, 10 videos and 10 audio clips as references, multi-turn extension. In Jimeng AI and Doubao; also in the Runway API (seedance2_5).
Joint audio-video generation from text, image, audio and video references; up to 15 s.
Name the feeling your product video needs and get five camera moves ranked for that product, with how each one breaks, the guard against it and a paste-ready prompt.
Turn one scene idea into four clips that cut together: one action per clip, lengths your tool offers, every subject described in full, guards against the usual failures and a paste-ready prompt per clip.
Write one Veo clip where people talk: lines cut to fit the clip length, a clear speaker for every line, timed beats and layered sound, plus checks for the result.
One product, ten different ad ideas checked against what AI video can render, ranked, and the winner written as ready prompts.
A matching b-roll library sized to your voiceover, with one look, audio that stays out of the way and a map from script lines to clips.
A short brand film with a real story: style lock, scene prompts sized to your model, a sound plan and a test-first generation order.
Camera moves chosen by the feeling each story beat needs, arranged so the energy builds on purpose, with prompts placed the way your model reads them.
One frozen character description, a reference setup for your model and a drift check per clip, so the same person appears in every scene.
A two-person conversation as Veo shots with native audio: lines sized to each clip, shot-reverse-shot coverage, steady voices and room tone.
A character's emotional turn in Kling, written as visible cues, kept small, and staged as a timed long take or a short multi-shot.
A lookbook film in Runway: the same model across looks, fabric that moves like fabric, and colourways edited from one approved shot.
Runway image-to-video planned frame first: keyframe prompts with one style block, motion-only prompts for Gen-4.5, and a clear rule for which to fix.
Background swaps in Runway Aleph 2.0 that match the subject's light, previewed as a still first, with a checklist that catches what makes a composite look fake.
Product, food and fashion shots that read as photographed: believable imperfections, motivated light, lens character, and a check against the real product.
Five openers for a product ad that move from frame one, keep the product recognisable, and hand off cleanly into the rest of the ad.
Transitions planned inside the clips: paired prompts with matched direction and speed, a marked cut point, and each model's tools for joining shots.
Test eight to ten settings cheaply, score them on how a person would look there, and turn the winner into a brief that keeps every later shot in the same place.
Slow, quiet sequences that make a product feel expensive: four restrained moves, long enough clips, empty space, one light, and sound before music.
Action prompts for Kling that describe weight, contact and material behaviour, with a checklist and a fix for each way the motion usually fails.
A whole mini scene as one Kling or Seedance generation: shots that fit the time, characters defined once, dialogue in the model's markup and a check per shot.
A song mapped to Runway clips: a look per section, clip lengths counted in bars, motion that is easy to cut on the beat, and the arithmetic checked.
Five product shots whose materials behave like the real thing, with a reference plan, a failure to watch per shot and two cut lengths.
A product rotation in Kling planned around what your photos show, with material wording, a lighting lock, a five-angle set and a check against the real product.
Make only chosen parts of a still move in Runway Gen-4.5, with a stillness list that keeps faces and walls put and fixes for the usual failures.
Your script or storyboard as a shot plan with film grammar, continuity, sound and paste-ready prompts, so separate clips cut into one scene.
One five-line style block, checked on three different subjects and tightened from your feedback, so a whole series looks like one director made it.
A 9:16 ad for TikTok, Reels or Shorts in Kling with three testable hooks, one shared body, safe-zone framing, honest claims and a test plan.