Top 30 AI Video Generators for Marketing in 2026

Last updated: August 2026

The AI video category spent 2024 and 2025 selling a promise. In 2026 it started sending invoices, and taking casualties.

The clearest signal came in March, when OpenAI announced it was discontinuing Sora less than six months after its public launch. The web and mobile apps went dark on April 26, 2026, and the API is scheduled to shut down on September 24, 2026.

Disney walked away from a three-year licensing agreement and a reported $1 billion stake in the process. Reporting from The Wall Street Journal and TechCrunch pointed at an unglamorous cause: worldwide usage peaked around a million users, then fell by more than half, while compute costs stayed brutal.

For marketers, the lesson is not that AI video is overhyped. Adoption data says the opposite. Wyzowl's 2026 survey found 63% of video marketers have used AI tools to create or edit marketing videos, up from 51% the year before, the fastest-moving metric in their twelve-year survey.

The IAB reports US digital video ad spend on track for roughly $82 billion in 2026, with the large majority of buyers using or planning to use generative AI for video creative.

The lesson is that the platform you standardize on can disappear. Build a stack, not a dependency.

This list covers 30 tools across the five jobs marketing teams actually hire AI video for: cinematic generation, presenter and avatar video, performance ad creative, repurposing long-form into short-form, and high-volume template production.

Pricing reflects public information as of August 2026 and moves constantly. Verify before you enter a card.

How to read this list

Most "best AI video generator" roundups fail because they compare a text-to-video research model against a caption tool. Those are not competitors. They are different departments. The five categories below map to five different budget lines.

Most teams running real video programs end up paying for one tool in three or four of these rows. That is normal, not a failure of tool selection.

Cinematic and generative video models (1 to 8)

These generate original footage from text or images. They are where the quality race is being fought, and where the leaderboard changed three times in the first half of 2026.

1. Google Veo 3.1 (via Flow)

Best for: Brand film, ads with dialogue, anything needing sound baked in.

Veo 3.1 remains the only major model producing lip-synced spoken dialogue and environmental audio in a single generation pass. That is a meaningful advantage when your ad script depends on someone actually saying something. Google folded Flow, Whisk, and ImageFX into one workspace in February 2026, so mood-boarding, stills, and animation now live in the same place, with Nano Banana Pro handling the image layer.

Pricing: Free daily credits. Google AI Pro around $29/mo. AI Ultra at roughly $200/mo for 4K upscaling and higher limits.

Watch out: Credits burn per generation attempt, not per usable output. Brand packaging and faces still drift during motion, which matters if you are shooting a specific product.

2. Kling 3.0

Best for: Photorealistic humans in 9:16 social video.

Kuaishou shipped Kling 3.0 in early February 2026 as the first model with native 4K at 60fps, plus a storyboard tool for per-shot camera and pacing control and lip-synced audio in the same pipeline. Multiple independent comparisons put it at or near the top on cost per usable clip, which is the metric that actually matters when you are generating forty variations.

Pricing: Usable free tier with daily credit resets. Paid plans from roughly $10/mo, and among the cheapest per-second API rates in the category.

Watch out: Free tier capacity gets throttled during peak hours.

3. Runway Gen-4.5

Best for: Teams that need reproducible shots and an approval process, not lucky generations.

Runway's edge was never raw benchmark scores, though Gen-4.5 held the top spot on the Artificial Analysis text-to-video leaderboard through early 2026. It is the control surface: motion brushes, camera paths, Gen-4 References for character consistency, Act-Two for performance capture, and Aleph for editing footage that already exists via text prompt. A May 2026 update added native audio and vertical, subtitle-ready social templates.

Pricing: From $12/mo (Standard), credit-based, with Gen-4 Turbo consuming roughly half the credits of Gen-4.5.

Watch out: Steepest learning curve on this list. Budget real onboarding time.

4. Seedance 2.0 (ByteDance)

Best for: Campaigns needing the same character or product across many shots.

Launched February 8, 2026, Seedance's differentiator is multimodal reference input. It accepts a stack of images, video clips, and audio tracks simultaneously, reported at up to nine images, three clips, and three audio files, giving you compositional control that pure text prompting cannot reach. Several hands-on reviews rate it best in class for character consistency with the least cleanup required.

Pricing: Free tier through various access points, with paid tiers up to roughly $167/mo.

Watch out: This is the access problem on the list. ByteDance paused international API access in March 2026 during copyright discussions with Hollywood studios. International users largely route through third-party resellers, which is a real procurement and compliance question for a brand. It also caps at 1080p native.

5. Luma Ray3 / Dream Machine

Best for: Product and environment shots where materials have to behave correctly, and any pipeline with a colorist in it.

Ray3 was the first video model to generate native 16-bit HDR with EXR export, which means output drops straight into a professional color workflow instead of getting graded around. Later Ray3 revisions leaned hard into physical simulation: liquid on glass, fabric in wind, gravity. That is exactly the failure mode that ruins product footage.

Pricing: Free tier available. Credit-based paid plans, with full-quality HDR clips costing significantly more credits.

Watch out: Version churn has been fast enough that most third-party reviews are testing an older Ray build than the one you will get.

6. Pika 2.5

Best for: Stylized, effect-driven short-form on a small budget.

Pika is not trying to win the realism war. Pikaffects, Pikaswaps, and Pikaframes (upload a start image and an end image, and Pika generates the transition) are built for scroll-stopping social content where "obviously stylized" is the point rather than a defect.

Pricing: From roughly $8/mo, the cheapest serious generator here.

Watch out: Raw generation quality trails Kling, Veo, and Seedance on most public benchmarks. Do not use it for hero brand work.

7. Adobe Firefly Video

Best for: Regulated industries, large enterprises, and any legal team that asks where the training data came from.

Firefly's native model is trained on licensed Adobe Stock content, which gives it the cleanest commercial-rights story in the category, with enterprise indemnity available on eligible plans. It also functions as a multi-model hub, since Veo and other partner models are selectable inside the same interface, and it lives inside Creative Cloud where your editors already work.

Pricing: From around $10/mo standalone.

Watch out: The indemnity story applies to Adobe's own model. When you switch to a partner model inside Firefly, the IP terms change with it. Read the per-model terms.

8. Higgsfield

Best for: Creative teams that want to compare looks across models without buying five subscriptions.

Higgsfield's pitch is aggregation: a large roster of video models behind a single subscription, plus presets and motion controls layered on top. For agencies pitching multiple creative directions, testing the same prompt across Kling, Veo, and others in one session is a genuine time saver.

Pricing: Subscription tiers with credit-based consumption across models.

Watch out: Aggregators sit downstream of provider outages, price changes, and regional restrictions. You inherit every upstream problem.

Avatar and presenter platforms (9 to 14)

Talking-head video without a camera, a studio, or a person who quits. This is the most commercially mature corner of AI video. These tools were profitable before the cinematic models were usable.

9. Synthesia

Best for: Enterprise-scale presenter video with procurement, compliance, and an LMS attached.

Synthesia offers roughly 240 avatars across 160+ languages, plus SCORM export, SSO, screen recording, and slide-to-video. It sells minutes rather than credits, which makes budgeting unusually predictable, since you know what a finished minute costs before you make it. For a marketing team producing localized product explainers on a schedule, that predictability is worth more than a few points of avatar realism.

Pricing: Free plan with 10 minutes/month. Starter around $18 to $29/mo, Creator around $64 to $89/mo, enterprise per-seat.

Watch out: Studio-grade custom avatars require a professional recording session and carry a significant annual fee.

10. HeyGen

Best for: Marketing and social teams that want the most lifelike presenter, and anyone localizing existing footage.

HeyGen's Avatar IV generation is widely rated the most expressive in the category, and its Video Translation feature, which redubs footage you already shot into 175+ languages with matched lip movement, has no direct equivalent at Synthesia. Instant avatars from a selfie take about five minutes. There is also a Streaming Avatar API for real-time and interactive use cases.

Pricing: Free plan with 3 videos/month. Creator around $29/mo, Business around $149/mo.

Watch out: Credit-based pricing means premium avatar engines consume credits faster. Collaboration features sit on the Business tier, which is a fivefold jump from Creator.

11. Colossyan

Best for: Sales enablement, onboarding, and training content that needs to be interactive.

Colossyan is narrower than HeyGen or Synthesia by design, since workplace learning is the whole thesis. What it adds that the other two do not is branching scenarios and embedded quizzes, which turns a passive explainer into something you can actually measure. Auto-translation covers roughly 31 languages.

Pricing: Free trial plus individual, team, and enterprise tiers.

Watch out: Full avatar customization and unlimited auto-translation are enterprise-gated.

12. Captions (Mirage)

Best for: Mobile-first creators and social teams who want polished talking-head output at a low entry price.

Now branded Mirage, the platform combines realistic AI avatars and digital twins with the multilingual editing and caption tooling it was originally known for. It publishes its pricing openly, which is more than several competitors in the ad-creative tier manage.

Pricing: Free tier, with paid plans from roughly $25/mo.

Watch out: Listed prices often reflect iOS plan pricing. Confirm web and desktop rates.

13. D-ID

Best for: Animating a single headshot, including executive photos, historical figures, and personalization at scale.

D-ID's specialty is turning any still portrait into a speaking avatar, which is a different job from building a reusable digital twin. It is the practical pick for personalized outbound video and for animating people who will not sit for a recording session.

Pricing: Free trial with credit-based tiers.

Watch out: Output quality is highly dependent on source photo quality. Get a good headshot or do not bother.

14. Vyond

Best for: Animated explainers where you do not want photorealism.

Vyond is the veteran here and the only entry built primarily around character animation rather than synthetic humans. For B2B explainers, internal comms, and any concept that is easier to draw than to film, such as abstract software workflows, org changes, and process diagrams, animation outperforms a generated actor standing in a generated office.

Pricing: Free trial with watermarked output. Starter around $99/mo, less annually.

Watch out: Priced well above the avatar tools. Justify it with volume.

Ad and UGC creative platforms (15 to 20)

Purpose-built for paid social. The unit of value is not one beautiful video. It is forty variants tested against each other by Thursday.

15. Creatify

Best for: E-commerce and DTC brands with a physical product that has to appear on screen.

Creatify's differentiator is genuinely putting your product in frame from a photo or product URL, rather than having a synthetic actor gesture at nothing. It publishes clear pricing, offers a free tier, and layers on AI media-buying and competitor tracking to scale what wins. Third-party analyses have put its effective cost around $4 per 30-second video on paid tiers.

Pricing: Free tier. Starter around $39/mo, Pro around $99/mo, higher tiers for unlimited output and analytics.

Watch out: Avatar realism is a step behind Arcads on close inspection, and free-tier output is watermarked.

16. Arcads

Best for: High-volume talking-actor testimonials for digital products such as SaaS, apps, courses, and coaching.

Arcads has the most convincing AI actors in the UGC category, with a library of 300+ digital performers and micro-expression quality that holds up better than its competitors on a second viewing. If you are selling something with no physical object to demo, that realism is the entire product.

Pricing: Not published. The pricing page returns a 404 and plans only appear after account creation. Third-party estimates put Starter around $110/mo for 10 videos and Creator around $220/mo for 20, or roughly $11 per video.

Watch out: Highest entry cost in the category, no free trial, and small edits require regenerating the whole clip at full credit cost. Verify the real number on the billing screen before committing.

17. AdCreative.ai

Best for: Teams that want creative scored before it ships.

AdCreative generates video and static ad creative and predicts performance against historical data, which shifts the workflow from "make everything, test everything" to "make everything, launch the plausible ones." For teams with limited testing budget, pre-launch scoring meaningfully reduces waste.

Pricing: Tiered subscription scaling with output volume and brand count.

Watch out: Predictive scores are directional, not gospel. Treat them as a prioritization signal, not a verdict.

18. Pencil

Best for: Brand teams that need AI creative to survive a brand-guidelines review.

Pencil sits at the more governed end of the ad-creative tier, generating video and static variants with predictive performance modeling while keeping brand assets and rules in the loop. It has increasingly positioned itself as a model aggregator with enterprise governance attached: role-based access, data ringfencing, and brand libraries. It is the option for organizations where "the AI made it" is not an acceptable answer to a legal question.

Pricing: Subscription tiers by team size and volume.

Watch out: Less synthetic-actor depth than the UGC specialists. Different tool for a different brief.

19. Icon

Best for: Brands that concluded AI UGC was not working and want the workflow with real humans.

Icon raised heavily to build "the AI Admaker" and then made one of the more interesting pivots of 2026. It now sells human-filmed UGC alongside its AI tooling, currently around $1,000 for a package of six ads filmed by vetted creators, with scripts, editing, and a creative director attached. The underlying Admaker software still handles AI generation, scene tagging, and ad tracking. Whatever you make of the positioning, the hybrid model, meaning AI for hooks and variants and humans for the body of the ad, is where a lot of serious spend is landing.

Pricing: Around $1,000 for the human-filmed UGC package. Software tiers priced separately.

Watch out: An order of magnitude above every other tool here on a per-deliverable basis. It is a production service with software attached, not software.

20. Pippit

Best for: Turning a product URL into a finished ad, inside the ByteDance ecosystem.

Pippit is ByteDance's CapCut-powered creative agent, converting product links into ad-ready video with generated avatars, voiceover, and templates tuned for short-form native formats. Its 2026 build routes through Seedance for generation. Being downstream of TikTok means it understands what performs on short-form platforms better than most Western-built tools.

Pricing: Free tier with affordable paid tiers.

Watch out: Output can read as template-derived if you do not customize. As with any ByteDance product, factor platform-policy risk into a long-term bet.

Repurposing and clipping tools (21 to 26)

The highest-ROI category on this list, and the most underrated. You already have footage. These tools turn one webinar into a month of short-form.

21. OpusClip

Best for: Mining long recordings for clips worth posting.

OpusClip solved the genuinely hard part of repurposing: identifying which five minutes of a ninety-minute recording deserve attention. Every clip gets a virality score for triage, and ClipAnything lets you search footage semantically, describing a moment instead of hoping the algorithm surfaces it.

Pricing: Free tier with watermarks. Starter around $15/mo for 150 processing minutes, Pro around $29/mo.

Watch out: The editor is Pro-gated. On Free and Starter you download whatever the AI produced, mistakes included. Budget for Pro or pick something else.

22. Submagic

Best for: Making clips you have already cut look professionally edited.

Submagic began as a caption tool and its animated caption templates are still the best in the category. The auto-zooms, sound effects, and transitions do most of the work that separates a raw clip from something that holds attention. Translation and dubbing cover 100+ languages.

Pricing: From roughly $19 to $20/mo, less annually.

Watch out: Weaker AI moment-selection than OpusClip. It is a finishing tool, not a discovery tool, so most teams pair it with a clipper rather than replacing one.

23. Descript

Best for: Teams that want to edit video by editing a transcript.

Descript's text-based editing remains the fastest way to cut a talking-head video for people who write faster than they scrub timelines. Overdub, filler-word removal, and studio-sound cleanup handle the tedious 80%. Transcription covers 25 languages, translation 61, and dubbing 30.

Pricing: From roughly $16 to $24/mo, with Business around $50/mo.

Watch out: Dubbing is gated to the Business tier, and rendering slows noticeably on long projects.

24. Vizard

Best for: Marketing teams clipping high volumes of webinars and podcasts with shared brand standards.

Vizard leans into team workflow with shared workspaces, brand kits, custom fonts, and transcript-based editing that is available on entry tiers rather than locked behind an upgrade. API access starts at the Creator tier.

Pricing: Creator around $29/mo, Business around $39/mo, roughly half that annually.

Watch out: One credit per minute of source video, billed on upload length regardless of how many clips you keep. A two-hour webinar costs 120 credits whether you export three clips or thirty. Credits do not roll over.

25. Riverside

Best for: Teams that record the interview and clip it in the same place.

Riverside records studio-quality local video for remote interviews and podcasts, then applies AI clipping, captions, and transcripts to what it just captured. For a B2B content team running a customer interview series, collapsing recording and repurposing into one tool removes an entire handoff and an entire subscription.

Pricing: From around $15/mo.

Watch out: Clip detection is competent rather than best in class. If mining archives is your main job, a dedicated clipper still wins.

26. Kapwing

Best for: Browser-based collaborative editing when half your team is not a video editor.

Kapwing is the generalist of this group, covering clipping, captions, resizing, subtitling, and real-time collaboration in a browser tab with no install. For a marketing team where the social manager, the designer, and the intern all need to touch the same file, that matters more than best-in-class anything.

Pricing: Pro around $24/mo.

Watch out: Jack of all trades. It will not beat a specialist at any single job.

Template and text-to-video production (27 to 30)

Volume tools. Not precious, not cinematic, but they will get thirty on-brand videos out the door this month.

27. InVideo AI

Best for: Polished social ads and explainers assembled rather than generated.

InVideo takes a fundamentally different approach from the models at the top of this list. Instead of generating original pixels, it assembles video from 6,000+ templates and a library of 16M+ stock clips with AI voiceover and automated editing. For a lot of marketing briefs, assembly produces a more professional and more predictable result than generation, and nobody has to explain a six-fingered hand to the client.

Pricing: Free tier with watermarks, then paid tiers by export volume.

Watch out: Generation can be slow and occasionally unstable, and template-derived output can look template-derived. Product representation sometimes needs manual correction.

28. Canva

Best for: Teams already standardized on Canva who need video to match everything else.

The case for Canva is not that its video AI beats the specialists, because it does not. It is that brand kits, templates, approvals, and your existing design library are all already there, so a social video inherits the brand automatically instead of by discipline. For most small marketing teams, that integration beats a better tool nobody adopts.

Pricing: Free tier, plus Pro and Teams tiers.

Watch out: Generation quality trails dedicated models by a wide margin. Use it for assembly and light generation, not hero content.

29. CapCut

Best for: Free, fast, short-form editing that natively understands what performs.

CapCut remains the most-used short-form editor for a reason: a full editing suite with effects, captions, transitions, and templates at no cost, tuned by proximity to TikTok. It has no AI moment-detection, so you are selecting clips manually, but for teams already living in it, adding a clipper on top beats replacing it.

Pricing: Free, with a Pro tier available.

Watch out: Manual clip selection, plus the same platform-risk caveat as Pippit.

30. Pictory

Best for: Turning content you have already published into video.

Pictory inverts the blank-canvas problem. You bring a blog post, a URL, a slide deck, product photos, or a long recording, and it returns a video with stock visuals, AI voiceover, captions, and transitions. For content teams sitting on three years of blog archive, it is the cheapest video pipeline available.

Pricing: Free trial, then paid tiers by video volume.

Watch out: No AI avatars or characters. If you need a presenter, this is not the tool.

The full comparison table

Honorable mentions

• Wan 2.6 from Alibaba. Earlier Wan releases were open-sourced under Apache 2.0, and if you have GPU capacity or an engineer, per-clip cost drops toward zero. Wan 2.6 adds 1080p output up to 15 seconds with synchronized audio. Note that reporting on whether 2.6 itself is fully open-source is inconsistent, so verify the license before you build on it.
• Hailuo AI. Strong on surreal and atmospheric output, with one of the more generous free tiers. Good for brand work where being memorable beats being realistic.
• Grok Imagine. Notably strong composition and lighting for a model that is not marketed at marketers. Distributed through xAI's API and via Higgsfield.
• Lumen5, Fliki, and VEED. Reliable text-to-video workhorses. All three are competent, but none is currently differentiated enough to displace anything in the top 30.
• Tavus. Real-time and personalized video at scale. Worth a look for lifecycle and outbound use cases.

How to actually pick

Start from the bottleneck, not the tool.

If your bottleneck is original footage you cannot shoot, meaning a product in an environment you cannot access or a scenario you cannot stage, you need a cinematic model. Start with Veo 3.1 if your script has dialogue, Kling 3.0 if it is photoreal humans in vertical, and Runway Gen-4.5 if art direction has to be repeatable.

If your bottleneck is a person on camera, with no spokesperson, no studio time, and thirty markets to localize, you need an avatar platform. HeyGen for marketing polish, Synthesia for enterprise governance, Colossyan if it is training.

If your bottleneck is creative volume for paid social, where your ads fatigue faster than you can replace them, you need an ad platform. Creatify if the product must appear on screen, Arcads if it is a digital product and you can absorb the price.

If your bottleneck is distribution, meaning you make good long-form and post it once, buy a clipper first. This is the highest-ROI purchase on this list and the one teams delay longest.

And if your bottleneck is headcount, template tools are the honest answer. Canva or InVideo shipping thirty adequate videos beats Runway shipping two beautiful ones nobody sees.

Three things to check before you commit

1. Read the commercial-rights terms for your specific plan and model. Free tiers almost universally exclude commercial use. Multi-model platforms can have different IP terms per model inside the same interface. Adobe Firefly has the clearest story here. Most aggregators have the murkiest.
2. Calculate cost per usable video, not headline price. Credits burn on failed attempts. Some tools bill on source length regardless of output. Some gate the editor behind a tier upgrade, which means the entry plan produces content you cannot fix. Normalize everything to one finished 30-second video on the tier you would actually buy.
3. Assume any single tool can disappear. That is the Sora lesson, and it is not hypothetical. ByteDance also paused international API access to Seedance in March 2026 over copyright discussions. Keep source assets and scripts in systems you control, prefer tools that export standard formats, and do not build an irreplaceable workflow on one vendor's roadmap.

FAQ

Is AI-generated video safe to use commercially?

On paid plans, generally yes, but it depends on the tool and the specific model. Adobe Firefly's native model trains only on licensed content and offers enterprise indemnity on eligible plans, which is the strongest position available. Veo, Runway, Synthesia, and HeyGen all include commercial rights on paid tiers. Free tiers usually do not. Read the terms for the exact plan you buy.

Do I need to disclose AI-generated video in ads?

Requirements vary by jurisdiction and platform, and they are tightening. Major platforms increasingly require labeling for synthetic media, particularly anything depicting a realistic person. Check current platform policies and your local advertising regulations. When a synthetic human is endorsing your product, disclosure is worth doing regardless of whether it is mandatory.

What replaced Sora?

No single successor. Veo 3.1 absorbed most of the cinematic-with-audio use case, Kling 3.0 took the value-per-clip position, and Runway Gen-4.5 holds the professional control-surface niche. OpenAI has signaled that video capability will continue inside ChatGPT rather than as a standalone product. If you are still running anything against the Sora API, migrate before September 24, 2026.

How much should a marketing team budget for AI video?

A functional small-team stack, meaning one clipper, one avatar tool, and one generator, runs roughly $60 to $100 per month. Add ad-creative tooling and you are at $150 to $350 per month. Enterprise deployments with per-seat avatar licensing and governance requirements run into five figures annually. Against traditional production costs, the payback math is rarely the hard part. Adoption is.

Will AI video replace video production teams?

It has not so far. Wistia's research indicates AI's impact concentrates in mid-process tasks such as script drafts, subtitles, repurposing, and asset tagging, while briefs and final deliverables stay human-led. The teams getting the most out of these tools use them to increase the number of ideas they can test, not to remove people from the process.

Sources

• OpenAI Help Center, What to know about the Sora discontinuation: https://help.openai.com/en/articles/20001152-what-to-know-about-the-sora-discontinuation
• TechCrunch, Why OpenAI really shut down Sora (March 29, 2026): https://techcrunch.com/2026/03/29/why-openai-really-shut-down-sora/
• The Decoder, OpenAI sets two-stage Sora shutdown: https://the-decoder.com/openai-sets-two-stage-sora-shutdown-with-app-closing-april-2026-and-api-following-in-september/
• eMarketer, OpenAI to discontinue Sora app as Disney exits partnership: https://www.emarketer.com/content/openai-discontinue-sora-app-disney-exits-partnership
• IAB, 2026 Digital Video Ad Spend and Strategy Report: https://www.iab.com/insights/video-ad-spend-report-2026/
• Wyzowl 2026 State of Video Marketing, via Vidico: https://vidico.com/news/video-marketing-statistics/
• AI video adoption statistics roundup: https://adwave.com/resources/ai-video-statistics-2026
• Model comparison, Seedance 2.0 vs Kling 3.0 vs Veo 3.1: https://www.aifreeapi.com/en/posts/seedance-2-vs-kling-3-vs-sora-2-vs-veo-3
• Best video generation models 2026: https://pinggy.io/blog/best_video_generation_ai_models/
• Google Flow 2026 update guide: https://www.vidau.ai/google-flow-ai-update-2026-complete-guide/
• HeyGen vs Synthesia comparison: https://www.ngram.com/blog/heygen-vs-synthesia
• AI UGC tool comparison: https://www.hyperfx.ai/blog/best-ai-ugc-tools-for-ads-2026
• Arcads pricing analysis: https://www.eesel.ai/blog/arcads-ai-pricing
• AI clipping tools benchmarked: https://reap.video/reports/state-of-top-ai-video-clipping-tools-2026

Pricing and availability change frequently in this category. All figures reflect publicly available information as of August 2026 and should be verified with each vendor before purchase.