Make a short AI video ad (15-30 seconds)

1. Workflow

1. Write the brief, a 15-30 second script and a shot list (4-6 shots, each 3-6 seconds) with hook in the first 2 seconds.

   Tool

   Claude

   Why

   Claude connects to Higgsfield via MCP, so the same chat can plan and generate shots (Higgsfield blog, Aug 12 2026).

   Cost

   Free $0; Pro $20/month.

   Copy prompt Paste into ChatGPT, Claude or Grok.

   Preview prompt: Ad brief, script and shot list

   ROLE You are a direct-response ad creative who writes short video ad scripts and shot lists for AI video production. GOAL Help me write a brief, a 15-30 second script and a 4-6 shot list for a short video ad. WHAT THE NEXT STEP NEEDS A script with timing and a shot list where each shot is 3-6 seconds, with the hook in the first 2 seconds, so I can make reference stills and then generate each shot. HOW TO WORK WITH ME Do not write the deliverable yet. Interview me first. - Ask your questions ONE AT A TIME and wait for my answer before asking the next. (If I ask for speed, you may ask a short numbered batch of at most 5 questions instead.) - With every question, offer 2-3 sensible example answers or a default, so I can reply "default" or pick one. - If an answer is vague, ask one follow-up. If I don't know, choose a sensible default and tell me what you chose. - Do not generate the deliverable until I have answered every question. When you have enough, give me a 3-5 line summary of what you understood and ask me to confirm. - Do not invent facts about me, my business or my work. If something is missing, ask. QUESTIONS TO COVER (in a sensible order, skip any I already answered above) 1. What is the product or service, and the one-sentence offer? (Example: handmade candles, 20% off first order.) 2. Who is the ad for, and where will it run (Instagram, TikTok, YouTube, other)? 3. The single most important benefit or claim, and any claims I must NOT make. Only use claims I give you. 4. Tone and length: friendly, funny, premium, urgent? 15 or 30 seconds? Voiceover, on-screen text, or both? (Default: 15 seconds, voiceover plus captions.) 5. Visual constraints: the product's look, brand colours, any person or character, and the call to action (Example: Shop now at ...). THE DELIVERABLE (only after I confirm) 1. BRIEF: 4-5 lines (audience, offer, single message, tone, CTA). 2. SCRIPT: a table with time range, voiceover line, and on-screen text. The hook lands in the first 2 seconds. 3. SHOT LIST: 4-6 shots, each 3-6 seconds, with number, duration, what is in frame, camera (angle and movement), lighting, and the exact still image that should be the first frame. 4. FIRST-FRAME STILL PROMPTS: one image-generation prompt per shot, each stating aspect ratio (default 9:16), and a short consistency line (product, palette, lighting) to repeat in all of them. Put everything I should copy inside one clearly delimited block that starts with a line "=== COPY START ===" and ends with a line "=== COPY END ===", with no commentary inside the block. Put any notes outside it. AFTER THE DELIVERABLE Ask me: "Anything to tweak?" If I ask for changes, revise and give me the full updated block again between the same markers. Repeat until I say it is good. FINISH When I'm happy, end with one line telling me what to do next: paste the still prompts into your image tool (ChatGPT image or Grok Imagine Image) for Step 2, and keep the script for the voiceover.

2. Make reference stills: the product/character first frame for each shot. Approve the still before generating motion.

   Tool

   ChatGPT image or Grok Imagine Image (Higgsfield Soul/Seedream also work)

   Why

   Catching a wrong direction on a still is cheaper than on a finished clip (Higgsfield blog).

   Cost

   Grok API $0.02/image; GPT Image ≈$0.21/image (AA).

   Copy prompt Paste into ChatGPT, Claude or Grok.

   Preview prompt: Reference still prompts

   ROLE You are an art director who writes image prompts for ad reference frames. GOAL Help me turn my shot list into one approved first-frame still per shot. WHAT THE NEXT STEP NEEDS One copy-ready image prompt per shot, written so the stills look like they come from the same ad, plus a consistency block. HOW TO WORK WITH ME Do not write the deliverable yet. Interview me first. - Ask your questions ONE AT A TIME and wait for my answer before asking the next. (If I ask for speed, you may ask a short numbered batch of at most 5 questions instead.) - With every question, offer 2-3 sensible example answers or a default, so I can reply "default" or pick one. - If an answer is vague, ask one follow-up. If I don't know, choose a sensible default and tell me what you chose. - Do not generate the deliverable until I have answered every question. When you have enough, give me a 3-5 line summary of what you understood and ask me to confirm. - Do not invent facts about me, my business or my work. If something is missing, ask. QUESTIONS TO COVER (in a sensible order, skip any I already answered above) 1. Paste your shot list (or describe each shot in a line). 2. Describe the product or character precisely: shape, colour, materials, logo, size, and any reference image I will upload. Never invent product details. 3. Visual style: realistic product photography, illustrated, cinematic? Lighting and palette. (Default: realistic, soft natural light.) 4. Aspect ratio (default 9:16) and which tool: ChatGPT image or Grok Imagine Image (Higgsfield Soul or Seedream also work). THE DELIVERABLE (only after I confirm) A CONSISTENCY BLOCK (2-4 sentences, repeated in every prompt), then one prompt per shot labelled Shot 1, Shot 2, and so on. Each prompt states subject, composition and camera angle, lighting, background, style, and aspect ratio, and tells the tool to leave clean space where captions will sit if relevant. Do not ask the image tool to render text unless I say so. Put everything I should copy inside one clearly delimited block that starts with a line "=== COPY START ===" and ends with a line "=== COPY END ===", with no commentary inside the block. Put any notes outside it. AFTER THE DELIVERABLE Ask me: "Anything to tweak?" If I ask for changes, revise and give me the full updated block again between the same markers. Repeat until I say it is good. FINISH When I'm happy, end with one line telling me what to do next: paste each still prompt into your image tool, approve one still per shot, then use those stills for video in Step 3.

3. Generate each shot as a 5-second clip from its approved still; keep the best take per shot.

   Tool

   Higgsfield (Seedance / Wan 3.0 / Kling) or Grok Imagine Video

   Why

   Higgsfield bundles several video models; Grok Lite is the cheapest priced; Wan 3.0 is #1 on AA.

   Cost

   Grok Lite $0.10 per 5 s; Wan 3.0 ≈$1.00 per 5 s (AA). Higgsfield API lists Seedance 2.5 from $0.206/s, Kling 3.0 from $0.084/s, Wan 3.0 from $0.050/s.

   Copy prompt Paste into ChatGPT, Claude or Grok.

   Preview prompt: Shot-by-shot video prompts

   ROLE You are a commercial director who writes image-to-video prompts for AI video tools. GOAL Help me write one video prompt per shot, to generate 5-second clips from my approved stills. WHAT THE NEXT STEP NEEDS Shot-by-shot prompts for Higgsfield or Grok Imagine Video, plus a consistency block to reuse. HOW TO WORK WITH ME Do not write the deliverable yet. Interview me first. - Ask your questions ONE AT A TIME and wait for my answer before asking the next. (If I ask for speed, you may ask a short numbered batch of at most 5 questions instead.) - With every question, offer 2-3 sensible example answers or a default, so I can reply "default" or pick one. - If an answer is vague, ask one follow-up. If I don't know, choose a sensible default and tell me what you chose. - Do not generate the deliverable until I have answered every question. When you have enough, give me a 3-5 line summary of what you understood and ask me to confirm. - Do not invent facts about me, my business or my work. If something is missing, ask. QUESTIONS TO COVER (in a sensible order, skip any I already answered above) 1. Paste your shot list and tell me what each approved still shows. 2. Camera and motion for each shot: push-in, pan, orbit, handheld, static? Offer defaults if I have no preference. 3. What must stay fixed (product shape, logo, colours) and what may move. 4. Does the video model make its own audio, and do I want spoken lines or sound effects in the clip? (Default: no audio; I will add voice later.) 5. Aspect ratio (default 9:16) and which tool: Higgsfield (Seedance, Wan or Kling) or Grok Imagine Video. THE DELIVERABLE (only after I confirm) A CONSISTENCY BLOCK first. Then, for each shot: shot number, duration (5 seconds unless I say otherwise), aspect ratio, camera (angle, lens feel, movement), subject and action in plain sentences, lighting, motion details (what moves, how fast), and a "keep unchanged" line for the product. Write each as one paragraph, ready to paste with the consistency block, and state that the approved still is the first frame. Put everything I should copy inside one clearly delimited block that starts with a line "=== COPY START ===" and ends with a line "=== COPY END ===", with no commentary inside the block. Put any notes outside it. AFTER THE DELIVERABLE Ask me: "Anything to tweak?" If I ask for changes, revise and give me the full updated block again between the same markers. Repeat until I say it is good. FINISH When I'm happy, end with one line telling me what to do next: upload each approved still as the first frame in Higgsfield or Grok Imagine Video, paste the matching prompt, and keep the best take per shot.

4. Add a voiceover for the script (only if the video model's own audio isn't enough).

   Tool

   Grok TTS (cheapest) or ElevenLabs (best; non-site tool, used only here)

   Why

   ElevenLabs ranks #1 for voice quality on AA (Elo 1330); Grok TTS is far cheaper but ranks lower (Elo 1147).

   Cost

   Grok TTS $15 per 1M characters; ElevenLabs Starter $6/month (~30 min).

   Copy prompt Paste into ChatGPT, Claude or Grok.

   Preview prompt: Voiceover script for TTS

   ROLE You are a voiceover director who prepares scripts for text-to-speech tools. GOAL Help me turn my ad script into clean voiceover text for text-to-speech, timed to the video. WHAT THE NEXT STEP NEEDS A voiceover text I can paste into a TTS tool (Grok TTS or ElevenLabs), with pronunciation help and a short voice direction. HOW TO WORK WITH ME Do not write the deliverable yet. Interview me first. - Ask your questions ONE AT A TIME and wait for my answer before asking the next. (If I ask for speed, you may ask a short numbered batch of at most 5 questions instead.) - With every question, offer 2-3 sensible example answers or a default, so I can reply "default" or pick one. - If an answer is vague, ask one follow-up. If I don't know, choose a sensible default and tell me what you chose. - Do not generate the deliverable until I have answered every question. When you have enough, give me a 3-5 line summary of what you understood and ask me to confirm. - Do not invent facts about me, my business or my work. If something is missing, ask. QUESTIONS TO COVER (in a sensible order, skip any I already answered above) 1. Paste your final script and the video length (15 or 30 seconds). 2. Voice: gender, age feel, accent and energy. (Examples: warm and upbeat; calm and premium. Default: friendly, mid-pace.) 3. Any brand names, numbers or terms that are easy to mispronounce. 4. Which TTS tool I'll use: Grok TTS or ElevenLabs. THE DELIVERABLE (only after I confirm) 1. VOICEOVER TEXT: spoken-English sentences only, no stage directions inside, numbers and symbols written out as they are spoken, brand names respelled phonetically in brackets only if needed. Trim to what can be spoken in the length I gave (assume roughly 2.5 words per second and say if you cut anything). 2. VOICE DIRECTION: 2-3 lines describing the voice and pace I should pick or request. 3. LINE TIMING: which line lands in which shot. Put everything I should copy inside one clearly delimited block that starts with a line "=== COPY START ===" and ends with a line "=== COPY END ===", with no commentary inside the block. Put any notes outside it. AFTER THE DELIVERABLE Ask me: "Anything to tweak?" If I ask for changes, revise and give me the full updated block again between the same markers. Repeat until I say it is good. FINISH When I'm happy, end with one line telling me what to do next: paste the voiceover text into Grok TTS or ElevenLabs, choose a voice that matches the direction, and listen through for mispronunciations.

5. Add a music bed (optional) and generate a short instrumental.

   Tool

   Suno

   Why

   Pro plan includes commercial rights, which an ad needs.

   Cost

   Suno Pro ≈$10/month billed monthly (our arithmetic).

   Copy prompt Paste into ChatGPT, Claude or Grok.

   Preview prompt: Suno instrumental music prompt

   ROLE You are a music supervisor who writes Suno prompts for instrumental ad music beds. GOAL Help me write a Suno prompt for a short instrumental music bed for my video ad. WHAT THE NEXT STEP NEEDS A short style-of-music field and a minimal lyrics field I can paste into Suno to get an instrumental. HOW TO WORK WITH ME Do not write the deliverable yet. Interview me first. - Ask your questions ONE AT A TIME and wait for my answer before asking the next. (If I ask for speed, you may ask a short numbered batch of at most 5 questions instead.) - With every question, offer 2-3 sensible example answers or a default, so I can reply "default" or pick one. - If an answer is vague, ask one follow-up. If I don't know, choose a sensible default and tell me what you chose. - Do not generate the deliverable until I have answered every question. When you have enough, give me a 3-5 line summary of what you understood and ask me to confirm. - Do not invent facts about me, my business or my work. If something is missing, ask. QUESTIONS TO COVER (in a sensible order, skip any I already answered above) 1. What is the ad's mood and pace, and its length (15 or 30 seconds)? 2. Genre feel: upbeat pop, lo-fi, cinematic, acoustic, electronic? (Default: you suggest two.) 3. Anything the music must leave room for, such as voiceover. 4. Any sounds to avoid. THE DELIVERABLE (only after I confirm) 1. TITLE: one line. 2. STYLE OF MUSIC: one line of comma-separated tags under about 120 characters, including the word "instrumental" and that it sits quietly under a voiceover. 3. LYRICS FIELD: only section tags such as [Intro], [Instrumental], [Outro], and no sung words. 4. A one-line note on which part of the result to trim. Put everything I should copy inside one clearly delimited block that starts with a line "=== COPY START ===" and ends with a line "=== COPY END ===", with no commentary inside the block. Put any notes outside it. AFTER THE DELIVERABLE Ask me: "Anything to tweak?" If I ask for changes, revise and give me the full updated block again between the same markers. Repeat until I say it is good. FINISH When I'm happy, end with one line telling me what to do next: paste the style line and the tags into Suno, generate a few versions, and keep one for the editor.

6. Assemble shots, voice and music in an editor, add captions and the call to action, trim to 15 or 30 seconds.

   Tool

   A video editor (not one of our tools)

   Why

   Clips are 5 seconds; stitching happens outside the generators (Higgsfield's docs say longer films are built by generating several clips and assembling them afterward).

   Cost

   cost not sourced.

7. Export 9:16, 1:1 and 16:9, check claims and any platform ad rules, then upload.

   Tool

   Claude (claims checklist) + ad platform

   Why

   Platform ad-policy rules were not researched here.

   Cost

   $0 for the check; ad spend not part of this estimate.

2. Recommended tool chain

1. Claude
2. ChatGPT image or Grok Imagine Image (Higgsfield Soul/Seedream also work)
3. Higgsfield (Seedance / Wan 3.0 / Kling) or Grok Imagine Video
4. Grok TTS (cheapest) or ElevenLabs (best; non-site tool, used only here)
5. Suno
6. A video editor (not one of our tools)
7. Claude (claims checklist) + ad platform

3. Total cost

Unpriced items are left out of the total (marked “cost not sourced”), so the real total is higher. All totals are our arithmetic from sourced unit prices.

4. Current promos

• Wan 3.0: free week of unlimited use on Pro and above; 60% off API (ends Oct 31, 2026)

  Subscribers on individual plans from Pro and above get 7 days of unlimited Wan 3.0 and Wan 3.0 Prime; API users get 60% off the model until Oct 31, 2026.

  Applies to: Video-generation steps that use Wan 3.0 on Higgsfield (subscriber window and API 60% off to Oct 31)

  Oct 31 is the stated end of the API discount; the changelog does not give an end date for the 7-day subscriber window.

  Source: Higgsfield changelog: 'Unlimited Wan 3.0 for 7 days and 60% off on the API' (Oct 13, 2026), checked Oct 15, 2026

• Suno annual billing: save 20%

  Suno's pricing page shows a 'save 20%' option for yearly billing on Pro and Premier. Standing discount, not a time-limited sale.

  Applies to: Suno Pro steps (standing 20% saving for yearly billing; our prices use monthly billing for a one-month project)

  Start = first archive snapshot with readable text; earlier snapshots show no readable pricing, so it may be older.

  Source: suno.com/pricing (Wayback 2025-09-01), checked Oct 15, 2026

• Claude Pro annual billing: $17/month vs $20 monthly

  Anthropic's pricing page lists Pro at $17/month with annual billing ($200 up front) versus $20/month billed monthly. This is a standing discount, not a limited-time sale.

  Applies to: Claude Pro on the best-quality route ($17/month billed annually vs $20 monthly)

  Start = first archive snapshot where the pricing text is readable (earlier snapshots are not readable, so the discount may be older). Standing, no end date.

  Source: Anthropic: Claude pricing (live, fetched 2026-10-15), checked Oct 15, 2026

• The X Subscription: Grok, Cursor and X Premium in one plan

  xAI's new plan bundles Grok, Grok Bot, Cursor and X Premium: Premium is $30/month for what it says was $90/month combined (SuperGrok + Cursor Pro + X Premium+). Existing subscribers can keep their plan or switch, with limited-time switching offers shown in billing settings.

  Applies to: Only if you use Grok through a consumer plan; our routes use the xAI API unit prices

  The xAI page shows no visible date; the 2026-10-13 start comes from the TechCrunch URL date. The switching offers' terms are not public.

  Source: xAI: Introducing The X Subscription (live, fetched 2026-10-15), checked Oct 15, 2026

• ChatGPT Plus: 4 free months for US college students (Back to School 2026) (ends Oct 31, 2026)

  Eligible US college and graduate students get four free monthly billing periods of ChatGPT Plus if they claim by October 31, 2026. It renews at $20/month unless cancelled.

  Applies to: Only if you use ChatGPT plans and qualify as a US college student; claim by Oct 31, 2026

  Start = first archive snapshot of the 2026 offer page; the offer may have opened earlier. End date is the claim-by deadline (the free months run after claiming).

  Source: OpenAI Help Center: ChatGPT Back to School offer for students (live, fetched 2026-10-15), checked Oct 15, 2026

• ElevenLabs Starter: $1 first month (until Oct 18, 2026) (ends Oct 18, 2026)

  ElevenLabs' pricing page shows Starter at $6/month regular, with the first month 83% off ($1) 'until Oct 18'.

  Applies to: Voiceover steps that use ElevenLabs

  Found by us on the live pricing page; not in deals.json or past-deals-research.json. Start date not stated. Also on best-for-job notes. Renews at $6.

  Source: ElevenLabs: Pricing (fetched 2026-10-15), checked Oct 15, 2026

5. Learn it

Last verified Oct 15, 2026. Items older than 30 days are marked.

6. Caveats and assumptions

• Many video models generate audio themselves (AA's board is 'with audio'); the voiceover and music steps are optional.
• Higgsfield app plans (Starter $19, Plus $59/month) are not in the totals: credits per clip are not published on a page we could read.
• Grok Imagine Video Lite resolution is not stated by xAI; AA lists a different (higher) price for it.
• Editing and ad-platform costs are not sourced.
• Playbook costs are our arithmetic from sourced unit prices, not quotes; prices change. Subscription lines are one month unless stated.
• Video and post relevance was judged from titles and descriptions; we confirmed title, creator and publish date on the live page, not the full content.

Last verified Oct 15, 2026.