How to Create a Wedding Invitation with AI: Photo & Video

2026-09-10 · blog how-to · Higgsfield · original

Topic: Video generation & shot direction (also: Assets, references & consistency, Image generation & editing, Cost, credits & plans, Pre-production & story, Prompting & AI assistants, Audio, voice & music, Post-production & finishing) · mode: Tutorial · tags: flattened table, prompts shared

Higgsfield's blog of 10 September 2026 lays out a seven-step pipeline for an AI wedding invitation made from the couple's photos, ending in a film of up to 30 seconds in Cinema Studio 4.0. It moves from Soul ID characters and a timed scenario to a style block, recurring props as Elements, and key frames in matched image models, with lettering added after the film. Iterating on cheap stills and fixing failed moments in Edit Video rather than rerunning keeps one invitation at about 241 credits (about $12).

What is said

Credit cost of one invitation and answers to common questions

What is said

★ — how much it matters to the source.

★★★ Higgsfield's blog says an AI wedding invitation can be made from ordinary photos of the couple in seven steps, from character training to a finished film of up to 30 seconds in Cinema Studio 4.0. · Cinema Studio 4.0 · also in: AI Human Video Generator
★ To start an AI wedding invitation you need several clear, recent photos of each partner, the invitation details, a basic story concept, optional references (real outfits, favourite locations, a prop) and a visual direction.
★★★ For a wedding invitation, create a character for each partner in Soul ID from 20 to 80 photos of that person: clear, recent, with varied angles and at least one full-height shot. · Soul ID
★★ Soul ID training runs once; the character is then used through the Soul 2.0 model, appears in Elements automatically and is tagged with @ in prompts. · Soul ID, Soul 2.0, Elements
★★ After training the Soul ID characters, Higgsfield advises generating a few portraits to confirm the likeness and see how each partner reads in different light and framing. · Soul ID
★ The couple in Higgsfield's wedding-invitation example does not exist: their reference selfies were generated in Soul Cinema and then treated like real photos. · Soul Cinema
★★ To put the real wedding outfits on the couple, upload a photo and photos of the clothing to AI Stylist piece by piece; the results become outfit references for the key frames.
★★ Higgsfield says the number of references an image model supports depends on the model and is usually enough for both partners, their outfits and a location at once.
★★★ For a short set of frames Higgsfield advises using real photos of the couple directly as references; Soul ID training pays off when the same faces must hold across a large set of wedding visuals. · Soul ID
★★★ Higgsfield advises writing an invitation film's scenario before generating anything and splitting it into arcs with time ranges: a 25–30 second film holds 6 to 8 beats of 2 to 4 seconds each.
★★ Higgsfield's scenario template for an invitation film: opening (2–4 s) with one close-up, development (8–12 s) across two or three locations, pause (3–4 s) of quiet close-ups, culmination (4–5 s) with the couple together, finale (5–6 s) with the spoken invitation line and the save-the-date card.
★★★ Higgsfield advises writing a style block, a fixed phrase added to the end of every image prompt with the palette, light, texture and mood (for example "palette of ivory, champagne gold and powder blue … natural directional light, 35mm film texture").
★★ In Cinema Studio 4.0, pick the colour palette template closest to the style block from the 50+ built-in palettes, or create a custom one, so the frames and the film share one grade. · Cinema Studio 4.0
★★★ Create or save a recurring prop as an Element and reference it with @ wherever it appears, to keep the same object consistent across the project; the couple's trained characters are referenced the same way. · Elements
★★★ Higgsfield advises doing all iteration on the key frames, one still per beat, where images cost a fraction of a video generation: generate each beat, tag the couple and props with @, end every prompt with the style block, and regenerate until the frame is right. · keyframe
★★ Higgsfield says Seedream 5.0 Pro combines several reference photos of people into one shot with unified lighting. · Seedream 5.0 · first in: 2026-07-08, Changelog: Seedream 5.0 Pro. Your idea, rendered exactly
★★★ For wedding key frames Higgsfield matches models to jobs: GPT Image 2 for general-purpose frames and reliable text, Nano Banana Pro for precise object control, Seedream 5.0 Pro for shots with both partners, Soul Cinema for a film-style grade. · GPT Image 2, Nano Banana Pro, Seedream 5.0, Soul Cinema
★★ Frames generated in any image model can be passed through Soul Cinema for a cinematic pass, so the set reads as stills from one film. · Soul Cinema
★★ One key frame can double as the photo invitation: in Higgsfield's example the rooftop shot serves as the digital invitation, a printed card, a Reels or Stories cover and the background of the save-the-date card.
★★ Generate the names and date as a separate lettering asset in a model with reliable text rendering: put the exact wording in quotes, describe the typeface (a well-known one can be named) and use a plain neutral background.
★★★ For a card inside the video, keep the background frame blank and add the typography after the film is generated, because text in motion is regenerated on every video pass.
★★ The photo invitation takes three assets: the chosen key frame, the lettering, and their combination in one image generation (both images as references) or any editor.
★★★ Cinema Studio on higgsfield.ai loads version 4.0 by default; Cinema Studio 4.0 takes up to 50 references per generation (images, videos and characters) and produces clips up to 30 seconds. · Cinema Studio 4.0, Cinema Studio
★★ Cinema Studio 4.0's Director's Panel sets seven genres, an era from the 60s to the 2020s, a tempo from Single Shot to Chaotic, cameras, lenses, apertures from f/1.4 to f/11, camera moves, 50+ palettes, six lighting presets plus custom, and a character emotion slider. · Cinema Studio 4.0
★ For an invitation film Higgsfield recommends a Clean Sharp or Anamorphic lens and f/1.4 for intimate close-ups.
★★★ For the invitation film in Cinema Studio 4.0, select the approved key frames and the couple's characters as references, then turn the timed scenario into the prompt: every beat with its time range, the action, the spoken finale and the music arc. · Cinema Studio 4.0
★★ Cinema Studio 4.0 generates audio in the same pass as the video, including speech, music and sound effects, so the invitation line to camera is part of the single generation. · Cinema Studio 4.0, native audio
★★★ When one moment of the film fails, Higgsfield advises the built-in modes (Edit Video, Genjutsu) instead of a full rerun, keeping full regeneration as the last resort. · Genjutsu
★★ Edit Video makes regional edits, fixing an object or a face in one moment of a clip while aiming to preserve the rest of it.
★★ Genjutsu transfers motion from a real clip onto generated characters, so the couple's actual first-dance video can drive their movement in the film; it can also replace an object in a finished video using reference images. · Genjutsu, motion transfer
★ To finish, overlay the lettering on the film's closing frames or append the save-the-date card as a final still, and export the film in 16:9.
★★★ Higgsfield says frame-level iteration is the main cost lever, since regenerating an image costs a fraction of regenerating the film. · credit
★★★ Higgsfield puts one AI wedding invitation at about 241 credits (about $12), covering nine images and the film; iterations and optional stages add to it. · credit, Cinema Studio 4.0
★★ A 28-second 16:9 invitation film at 720p in Cinema Studio 4.0 costs 182 credits ($9.10). · Cinema Studio 4.0, credit
★★ Key frames, prop and lettering images at 2K cost 6.5 credits ($0.33) per image in Higgsfield's wedding-invitation pipeline. · credit
★★ In Higgsfield's wedding-invitation costs, Soul ID character training costs 25 credits ($1.25) per partner, an AI Stylist try-on 2 credits ($0.10) per generation, and a reference selfie 0.125 credits per image. · Soul ID, credit
★★ The minimal invitation path skips the optional stages: direct photo references instead of Soul ID training, and outfits described in prompts instead of AI Stylist. · Soul ID
★★ On plans where a model is in the Unlimited list, its generations on higgsfield.ai deduct no credits; the current list per plan is on the Pricing page. · Higgsfield Unlimited

Threads in this video

refine · How do you keep a character the same across shots? — For a short set of frames Higgsfield advises using real photos of the couple directly as references; Soul ID training pays off when the same faces must hold across a large set of wedding visuals.
refine · How do you put a new outfit on a person in a photo? — To put the real wedding outfits on the couple, upload a photo and photos of the clothing to AI Stylist piece by piece; the results become outfit references for the key frames.
refine · How do you write the script for an AI film? — Higgsfield advises writing an invitation film's scenario before generating anything and splitting it into arcs with time ranges: a 25–30 second film holds 6 to 8 beats of 2 to 4 seconds each.
refine · How do you keep a consistent colour palette and visual style across generated images? — Higgsfield advises writing a style block, a fixed phrase added to the end of every image prompt with the palette, light, texture and mood (for example "palette of ivory, champagne gold and powder blue … natural directional light, 35mm film texture").
refine · How do you keep a location, product or prop the same across shots? — Create or save a recurring prop as an Element and reference it with @ wherever it appears, to keep the same object consistent across the project; the couple's trained characters are referenced the same way.
refine · How do you draft and lock a shot before rendering the final? — Higgsfield advises doing all iteration on the key frames, one still per beat, where images cost a fraction of a video generation: generate each beat, tag the couple and props with @, end every prompt with the style block, and regenerate until the frame is right.
refine · Which image model should be used for keyframes, edits, text and outfits? — For wedding key frames Higgsfield matches models to jobs: GPT Image 2 for general-purpose frames and reliable text, Nano Banana Pro for precise object control, Seedream 5.0 Pro for shots with both partners, Soul Cinema for a film-style grade.
refine · How do you get a cinematic look in a generated image or shot? — Frames generated in any image model can be passed through Soul Cinema for a cinematic pass, so the set reads as stills from one film.
refine · How do you keep on-screen text and fonts intact in a generated video? — For a card inside the video, keep the background frame blank and add the typography after the film is generated, because text in motion is regenerated on every video pass.
refine · How do you direct a multi-shot sequence? — For the invitation film in Cinema Studio 4.0, select the approved key frames and the couple's characters as references, then turn the timed scenario into the prompt: every beat with its time range, the action, the spoken finale and the music arc.
refine · Should sound and music be generated with the clip or added in the edit? — Cinema Studio 4.0 generates audio in the same pass as the video, including speech, music and sound effects, so the invitation line to camera is part of the single generation.
refine · When a generation fails, should you generate more takes, fix the prompt or start over? — When one moment of the film fails, Higgsfield advises the built-in modes (Edit Video, Genjutsu) instead of a full rerun, keeping full regeneration as the last resort.
refine · How do you get precise motion (dance, action, choreography)? — Genjutsu transfers motion from a real clip onto generated characters, so the couple's actual first-dance video can drive their movement in the film; it can also replace an object in a finished video using reference images.
refine · How do you spend fewer credits? — Higgsfield says frame-level iteration is the main cost lever, since regenerating an image costs a fraction of regenerating the film.
refine · What does an AI clip, shot or film cost? — Higgsfield puts one AI wedding invitation at about 241 credits (about $12), covering nine images and the film; iterations and optional stages add to it.

Mentioned

Central: Cinema Studio 4.0 · Soul ID

Also: Concepts: keyframe · motion transfer · style prefix

Products: AI Stylist · Director's Panel · Elements · Genjutsu · GPT Image 2 · Higgsfield Unlimited · Nano Banana Pro · Seedream 5.0 · Soul 2.0 · Soul Cinema · Video Edit

← Previous: AI Motion Designer: Higgsfield Meets After Effects Through GPT · Next → Changelog: Effects 2.0. Viral presets in ChatGPT