AI Models & Providers

We choose models for quality, safety and price. Every tool shows its price before you start, and where a tool offers a choice of models, you pick the model and pay for the one you pick. The tables below have 64 entries; a few of them cover a family of related models.

Chosen per scene or clip in the video studios. Each tool shows the model and the price before you render.

Chosen in the Image Generator, the Thumbnail Generator and other image tools.

Used inside the photo tools and the design editor.

Used in the Voice Studio, the presenter and translation tools, and for narration.

Since October 2026 every stock voice in QikTools, and timed transcription in Brand Guardian, captions, translation and our newer tools, use these models instead of OpenAI's retiring speech models.

Used in the AI Music Generator, the Campaign Orchestrator and the video studios.

Used by the writing and business tools, the planners, the help assistant and our safety systems.

Used by Brand Models (private models trained on your own images) and product 3D models.

Each provider below makes or serves models you can use in QikTools. Its usage policy applies to the requests you send to its models, and its terms decide what it may do with the content it receives. Where a provider lets us switch off training on your content, we do; where it doesn’t, its own terms govern that use and we have no control over it. We describe a provider as not training on your content only when its own terms say so.

• Runway

  Directly from its developer

  May train on content

  Models

  GWM-1, Gen-4.5, Gen-4 Turbo, Gen-4 Image

  Used for

  Talking-head presenter videos (GWM-1); Video scenes and clips in the video studios (Gen-4.5, Gen-4 Turbo); Images that keep the same character or product (Gen-4 Image); Running other companies’ models for us: Seedance 2.5, Seedance 2.0 at 1080p, Seedream 5 Pro and ElevenLabs’ Expressive voice (Eleven v4), voice changer, dubbing and voice isolation (see each company) — and, since 10 October 2026, Nano Banana Pro (any size) and Nano Banana 2 (4K) only as the automatic backup when Fal can’t make a picture (see Google)

  Training

  Runway’s Terms of Use (updated 15 September 2026, read again on 10 October 2026) say “Inputs and Outputs may be used by the Company to train and improve its AI models” — this covers everything we send through Runway, including other companies’ models it runs for us and the Nano Banana pictures it makes as a backup. A request made with “Only providers that don’t train on my content” never goes to Runway.

  Attribution

  “Powered by Runway”, linked to runway.com, shown prominently on every screen that uses a model served through Runway (Runway’s Terms of Use).

  Its rules for your requests (14)

  • Nothing that depicts, facilitates or promotes child sexual abuse or the sexualisation of children, and nothing designed to impersonate a minor.
  • No talking-head avatars or characters based on the face or voice of anyone under 18, and none aimed at children.
  • No avatars of public figures or copyrighted characters without authorization.
  • Don’t use an image, video or audio of another person without their permission.
  • Don’t impersonate a person or organisation, or misrepresent an affiliation with one.
  • Nothing that compromises other people’s privacy.
  • No sexually explicit content or adult nudity, and no non-consensual intimate imagery.
  • No terrorism or violent extremism, graphic violence or gore, or content that incites violence.
  • No content that dehumanises or discriminates against people for a protected attribute, harasses, bullies or defames others, or promotes self-harm or disordered eating.
  • Don’t ask for work in the style of a known, living artist, or anything that may infringe intellectual property rights.
  • Don’t use it to defraud, scam or deliberately mislead others.
  • Talking-head presenters may not give medical, legal or financial advice or act as a therapist or counsellor.
  • Apps you build on our API that show results from Runway-served models must display “Powered by Runway” linked to runway.com.
  • Don’t use Runway-served results to build or train a competing AI product.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 10 October 2026

• OpenAI

  Directly from its developer

  No training by default

  Models

  GPT-4.1, GPT-4.1 mini, GPT-4.1 nano, GPT-4o, GPT-4o mini, GPT-5.6 Luna, GPT Image 2.5, omni-moderation-latest, text-embedding-3-small, GPT-Transcribe, GPT-Realtime 2.1 mini

  Used for

  Writing, scripts, plans, captions and document answers; Reading images, documents and receipts; GPT Image generation and editing; Safety checks on requests before they run; Descriptions and search in the asset library; Text-only transcription with GPT-Transcribe: the Quick transcript of meeting minutes, podcast and Video to Blog uploads, voice notes, and checks of uploaded recordings; The help assistant, in text and, with Talk to Munns, by voice (listening and speaking in real time with GPT-Realtime 2.1 mini, and GPT-Transcribe for its captions); Explainer videos: the script and scene plan (GPT-4.1), the scene pictures (GPT Image 2.5) and the placing of labels on them (GPT-4.1 mini); Checking each line of a Voice Vault dub, word for word, with GPT-Transcribe; Translating non-English ad scripts into English (GPT-4.1 nano) so the claim check can read them before anything is rendered or charged

  Training

  OpenAI’s Services Agreement (v. 1 January 2026): “OpenAI will not use Customer Content to develop or improve the Services, unless Customer explicitly agrees to such use.”

  Its rules for your requests (14)

  • Nothing that exploits, endangers or sexualises anyone under 18, including grooming and age-inappropriate content for minors.
  • No sexual violence or non-consensual intimate content.
  • No threats, harassment, intimidation or defamation, and nothing that promotes suicide or self-harm.
  • No terrorism or violence, and no help developing weapons.
  • Don’t use a real person’s likeness, including their photorealistic image or voice, without consent in ways that could confuse people about what’s real.
  • Don’t build facial recognition databases without consent, identify other people with real-time biometrics in public spaces, or score and profile other people from biometric data.
  • Don’t aggregate, monitor, profile or share other people’s private information without their authorization.
  • No political campaigning, lobbying, election interference or voter demobilisation.
  • Don’t use it to manipulate or deceive others, including for academic dishonesty.
  • No fraud, scams or other illegal activity.
  • No tailored legal, medical or other licensed advice without a licensed professional involved, and no automated high-stakes decisions about credit, employment, housing or education without human review.
  • If you publish writing made with these models, don’t mislead readers into thinking it was written entirely by a person.
  • Don’t bypass or circumvent safety measures or usage limits.
  • Don’t use outputs to develop AI models that compete with OpenAI.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 3 October 2026

• Google

  Through a licensed API providerPowered by Runway

  No training by default

  Models

  Veo 3.1, Veo 3.1 Fast, Veo 3.1 Lite, Gemini Omni Flash 1.1, Nano Banana 2.1, Nano Banana 2, Nano Banana Pro, Lyria 2

  Used for

  Video scenes and clips (Veo 3.1, Fast and Lite, and Gemini Omni Flash 1.1 — through fal); AI motion for the scenes of an explainer video (Veo 3.1 Lite, through fal); Image generation and editing, and character and place portraits (Nano Banana 2.1 and Nano Banana Pro — through fal since 10 October 2026; Runway’s Nano Banana Pro (any size) and Nano Banana 2 (4K) only as an automatic backup when fal can’t make the picture); Cinematic music (Lyria 2 — through fal)

  Training

  Google’s Gemini API Additional Terms (effective 23 March 2026, page updated 28 April 2026, read again on 10 October 2026) say that for paid services “Google doesn’t use your prompts … or responses to improve our products”; we reach Google’s models through fal, which pays for them (fal marks the Nano Banana models we use “Enterprise Ready”, checked 10 October 2026), and, for Nano Banana pictures only as an automatic backup, through Runway — whose own terms still let Runway train on what passes through it. A request made with “Only providers that don’t train on my content” is never moved to Runway.

  Its rules for your requests (14)

  • No child sexual abuse or exploitation content.
  • No sexually explicit or pornographic content, and no non-consensual intimate imagery.
  • No violent extremism, terrorism, hate speech, harassment, bullying or incitement to violence, and nothing that facilitates self-harm.
  • Nothing that facilitates illegal activity or the trade in illegal goods or services.
  • Don’t impersonate a real person, living or dead, without clear disclosure, in order to deceive.
  • Don’t track or monitor other people without their consent.
  • Don’t infringe intellectual property rights.
  • No scams, fraud, spam, phishing or malware.
  • No misleading claims about elections, health or other matters that could cause harm, and don’t falsely claim expertise in sensitive areas.
  • Don’t mislead anyone about where content came from by claiming it was made solely by a human.
  • No automated high-stakes decisions in areas like employment, health care, finance, housing, insurance or welfare without human oversight.
  • Not for clinical practice, medical advice, or anything that needs medical-device approval.
  • Don’t bypass safety filters or abuse protections.
  • Don’t use these models to develop competing AI models.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 11 October 2026

• ElevenLabs

  Through a licensed API providerPowered by Runway

  May train on content

  Models

  Eleven v4, Eleven v4 Turbo, Eleven v3, Text to Dialogue (Eleven v3), Eleven Turbo v2.5, Scribe v2, Sound Effects v2, Eleven Multilingual v2 (speech to speech), Dubbing, Voice Isolator

  Used for

  The Voice Studio’s Expressive voice (Eleven v4, through Runway, with Eleven v4 through fal as its backup) and multi-speaker dialogue (Eleven v3, through fal); Natural narration (Eleven v4 Turbo, with Turbo v2.5 as its backup) and expressive narration in our newer tools, through fal; Word-timed transcription with speaker labels (Scribe v2, through fal): the AI Video Clipper, Detailed transcripts, Brand Guardian, captions, translation and our newer tools; Pro sound effects (through fal); Voice changer, dubbing (the Audio Translator) and voice isolation (through Runway); The AI Video Translator's voices for the languages xAI doesn't speak (Eleven v4 Turbo, through fal)

  Training

  ElevenLabs’ Terms of Service (31 March 2026) let it use customer content “to improve the Services” unless the account holder opts out; the account holders for our requests are fal and Runway, not us, so the opt-out isn’t ours to set.

  Its rules for your requests (21)

  • Nothing that sexualises, exploits or endangers children, nothing designed to impersonate a minor, and no age-inappropriate material aimed at minors.
  • Don’t replicate another person’s voice without their consent or a legal right to do so.
  • Don’t present an AI voice in a way meant to deceive people about whether it was made with AI.
  • Don’t use a cloned or imitated voice to harass or harm the person it imitates.
  • Don’t impersonate political candidates or elected officials, even with their permission.
  • No voter suppression or misleading election information.
  • No audio from ElevenLabs’ models that advertises or markets alcohol, tobacco, drugs, prescription medicines, supplements, herbal remedies or medical devices (ElevenLabs requires its written approval).
  • Don’t use audio from ElevenLabs’ models to promote real-money gambling or payday loans.
  • No tailored financial, legal or health advice unless a qualified professional reviews it first and the use of AI is clearly disclosed.
  • No high-stakes automated decision-making that affects someone’s wellbeing.
  • No fraud, scams, unauthorised robocalls or spam, and no evading safeguards such as verification checks.
  • Nothing that helps trade illegal goods or services, such as illegal drugs or weapons.
  • No violent threats, extremism, hate, harassment, stalking or content promoting self-harm, suicide or eating disorders (clearly fictional work and news reporting excepted).
  • No health misinformation or other harmful misinformation.
  • No unauthorised access to systems, malware or surveillance tools.
  • Don’t infringe intellectual property or privacy rights.
  • Don’t sell, license or distribute sound effects made with the Pro (ElevenLabs) engine on their own — as files, samples or sound libraries; use them inside your own projects.
  • Don’t use outputs to train AI models or build products that compete with ElevenLabs.
  • Don’t remove notices or markings from audio made with ElevenLabs’ models.
  • Government bodies may not use the ElevenLabs-powered features without ElevenLabs’ written authorization.
  • Don’t use ElevenLabs-made audio for anything EU AI law classes as prohibited or high-risk.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 3 October 2026

• fal

  Directly from its developer

  No training by default

  Models

  FFmpeg API (compose), Image Editing (object removal, photo restoration, scene composition, style transfer), FLUX.1 [schnell], SeedVR2, DeepFilterNet 3, Demucs, Chatterbox, TRELLIS, Qwen-Image-2512, Z-Image-Turbo

  Used for

  Runs most of the other companies’ models listed here, and stores generated files; Open models it runs on its own servers: quick image drafts (FLUX.1 [schnell]), upscaling (SeedVR2), noise removal (DeepFilterNet 3), vocal separation (Demucs, which the AI Video Translator also uses to separate a video’s speech from its background), speech in a verified cloned voice (Chatterbox), draft 3D models (TRELLIS) and custom brand models (Qwen-Image-2512, Z-Image-Turbo); Photo edits: object removal, restoration, scene composition and style transfer; Joining the scenes of cloud film renders and translated videos with their soundtrack (FFmpeg compose)

  Training

  fal’s API Services Terms: “Company will not use Client Content to create, train, develop (directly or indirectly) Company’s products or services” — for models fal marks “Pending Enterprise Ready” its security commitments and data processing addendum don’t apply. We only use models fal marks Enterprise Ready; cloud film renders are prepared and checked on our own servers and joined with fal’s Enterprise Ready FFmpeg compose.

  Its rules for your requests (11)

  • No child sexual abuse material, grooming or sextortion content, and nothing that sexualises, exploits or harms minors.
  • No non-consensual intimate imagery, sexual violence or trafficking content.
  • No deepfakes or impersonation that use a person’s name, image or likeness without consent in a commercial, defamatory or harmful way.
  • No threats, harassment, hate speech or incitement to violence, and nothing that promotes self-harm or suicide.
  • No misinformation likely to harm public health, safety or election integrity.
  • No scams, fraud, defamation or spam.
  • Nothing that infringes copyright, trademarks or other intellectual property.
  • Don’t process other people’s biometric data or sensitive personal information without authorization.
  • Don’t reverse-engineer models, extract weights, or circumvent safety filters.
  • No malware or attacks on the service.
  • Don’t resell or offer raw access to the models through our service.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 10 October 2026

• ByteDance

  Powered by RunwayThrough a licensed API provider

  Not stated

  Models

  Seedance 2.5, Seedance 2.0, Seedance 2.0 Fast, Seedance 2.0 Mini, Seedream 5 Pro

  Used for

  Video scenes and clips (Seedance 2.5 and Seedance 2.0 at 1080p through Runway; Seedance 2.0, Fast and Mini through fal); Photoreal and product images (Seedream 5 Pro — through Runway)

  Training

  ByteDance’s BytePlus terms for these models are only readable when signed in, so we can’t confirm them; through Runway, Runway says it negotiates no-training terms with every model maker it runs (Runway’s own entry still applies).

  Its rules for your requests (2)

  • Use Seedance and Seedream results in your own or your clients’ content; don’t resell access to the models.
  • Don’t try to get around ByteDance’s check that refuses pictures showing real people’s likenesses.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 4 October 2026

• Kuaishou (Kling AI)

  Through a licensed API provider

  May train on content

  Models

  Kling O1, Kling O3, Kling 3.0

  Used for

  Video scenes, product clips and reels that keep characters and products consistent (through fal)

  Training

  Kling AI’s Terms of Service (21 April 2026) let it use content to “create, test, improve, train” its models; we reach Kling through fal’s partnership, whose terms with Kling aren’t public.

  Attribution

  Kling AI’s Terms of Service ask users to label outputs with the Kling AI brand (§4.5) and to get Kling’s written permission before using outputs commercially (§4.6). We reach Kling through fal, whose model pages say “Commercial use via fal partnership” and don’t mention branding; we name “Kling O1”, “Kling O3” and “Kling 3.0” wherever they’re offered and in each video’s Content Credentials.

  Its rules for your requests (8)

  • No sexually explicit, obscene, pornographic, hateful or inflammatory material, and nothing defamatory.
  • Don’t impersonate a person or organisation, or misrepresent an affiliation with one.
  • No harassment, threats, bullying or discrimination.
  • No political campaigning, advocacy or lobbying content.
  • No fake reviews.
  • Nothing that infringes copyright, trademarks or other intellectual property, or anyone’s privacy.
  • Don’t use Kling outputs about someone to decide on their credit, education, employment, housing, insurance, legal or medical matters.
  • Don’t use outputs to train AI models or build competing products.

  Terms (opens in a new tab)Checked 11 October 2026

• Alibaba Cloud

  Through a licensed API provider

  No training by default

  Models

  Wan 3

  Used for

  Low-cost video scenes and clips with reference pictures (through fal); AI motion for the scenes of an explainer video (Wan 3, through fal)

  Training

  Alibaba Cloud Model Studio (page updated 28 September 2026): it “will never use your data for model training”; we reach Wan 3 through fal’s partner API.

  Terms (opens in a new tab)Checked 4 October 2026

• MiniMax

  Through a licensed API provider

  Not stated

  Models

  MiniMax H3, MiniMax H3 Max, MiniMax Music 2.0

  Used for

  Low-cost video scenes with reference pictures (through fal); Making a MiniMax H3 shot longer — MiniMax H3 Max continues the finished clip without a cut (through fal); Songs with vocals (through fal)

  Training

  MiniMax’s API platform terms only load in a browser script and we couldn’t read them; we reach MiniMax through fal’s partner API.

  Terms (opens in a new tab)Checked 4 October 2026

• Black Forest Labs

  Through a licensed API provider

  May train on content

  Models

  FLUX.1 Kontext [pro], FLUX 3

  Used for

  Photo edits, character angles, product staging and sketch-to-image (through fal); FLUX.1 [schnell] drafts run on fal’s own servers (see fal); Long video scenes and clips of 5 to 20 seconds with their own sound and dialogue (FLUX 3, through fal)

  Training

  Black Forest Labs’ FLUX API Service Terms (4 August 2026, read again on 14 October 2026): “the Company may use Inputs and Outputs to train and improve its artificial intelligence models”; we reach FLUX.1 Kontext [pro] and FLUX 3 through fal’s partner API.

  Its rules for your requests (12)

  • Don’t realistically depict a real person in a sexual, intimate, degrading, defamatory, fraudulent or misleading way without their verified, documented, informed consent.
  • No child sexual abuse material or non-consensual intimate imagery, and nothing depicting minors in a sexual, violent or harmful way.
  • No disinformation or realistic content of real people meant to mislead voters, and no political campaigning, electoral advertising or voter suppression.
  • Don’t share content that falsely shows events or things in a way likely to deceive a reasonable viewer, without proper disclosure.
  • Don’t remove, alter or interfere with Content Credentials, watermarks or other provenance signals, or claim an output was made by a human.
  • No harassment, abuse, threats, stalking, defamation or bullying.
  • No unlawful impersonation or use of a real person’s name, image, voice or likeness, and no privacy violations.
  • Get documented consent from anyone whose name, image, likeness or voice you use, both for the upload and for how you’ll use the result.
  • No military, surveillance, law-enforcement or biometric-processing uses, and no use in a high-risk AI system.
  • No fully automated decisions that harm someone’s legal rights, and no unlawful discrimination in employment, finance, housing or education.
  • Nothing that infringes intellectual property or other third-party rights.
  • Don’t use outputs to train other AI models or build competing products.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 14 October 2026

• SpaceXAI (xAI)

  Through a licensed API provider

  No training by default

  Models

  xAI Text to Speech

  Used for

  Standard voices and narration across QikTools — the Voice Studio, voiceovers in the video studios and the API — in 20 languages including Hindi and Bengali (through fal); The AI Video Translator's matched voices (through fal)

  Training

  xAI’s API security FAQ: it “never trains on your API inputs or outputs without your explicit permission”; we reach it through fal’s partner API.

  Its rules for your requests (4)

  • Nothing that sexualises or exploits children.
  • Don’t deceptively impersonate a real person.
  • No help developing biological, chemical or other weapons of mass destruction.
  • Don’t jailbreak, circumvent safeguards or disrupt the service.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 4 October 2026

• Sync Labs (sync.)

  Through a licensed API provider

  May train on content

  Models

  lipsync-2, lipsync-1.9.0-beta

  Used for

  Lip-sync for presenters and film scenes, and the Video Translator's Close-up lip-sync (lipsync-2, through fal); lipsync-1.9.0-beta only finishes translations started before 5 October 2026

  Training

  Sync Labs’ Terms of Service (revised 6 June 2024) give it a licence to use uploaded content “to develop and improve the Service”; we reach it through fal’s partner API.

  Its rules for your requests (4)

  • Only lip-sync people whose consent you have, with the rights to use their image and likeness.
  • Don’t impersonate a person or organisation, or misrepresent an affiliation with one.
  • Nothing unlawful, threatening, abusive, harassing or excessively violent.
  • No spam or unsolicited advertising.

  Terms (opens in a new tab)Checked 4 October 2026

• VEED

  Through a licensed API provider

  No training by default

  Models

  VEED Lipsync

  Used for

  The Video Translator's Standard lip-sync (through fal) — for short videos Standard runs on Creatify Lipsync instead

  Training

  VEED’s API FAQ (read 7 October 2026): “We don’t use your API inputs or outputs to train our models unless you agree to it” (it keeps each request’s log for 7 days); its Terms of Use (22 April 2024) only let it train on content from free VEED accounts. Veed Limited is in London, United Kingdom; fal marks VEED Lipsync as a partner API, so we reach it through fal.

  Its rules for your requests (10)

  • Only lip-sync people whose consent you have, with the rights to use their image and likeness.
  • No deepfakes that falsely depict a person, no fraudulent endorsements and no misleading political content.
  • Nothing that promotes, advertises or endorses a political party, candidate or elected official.
  • No non-consensual sexual content, and no sexual content depicting anyone who appears to be under 18.
  • No bullying, stalking, harassment, hate or discrimination.
  • Nothing false that may create unjustified alarm.
  • No spam, phishing, fraud or unauthorised advertising.
  • Don’t infringe intellectual property rights or copy the style of a known artist.
  • Don’t upload government ID numbers, passwords, payment card or account details, health information or other special categories of personal data, or data about criminal convictions.
  • Don’t remove or obscure markings that identify content as AI-made.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 7 October 2026

• Creatify

  Through a licensed API provider

  May train on content

  Models

  Creatify Lipsync

  Used for

  The Video Translator's Premium lip-sync, and its Standard lip-sync for short videos (run by fal on its own servers)

  Training

  Creatify’s Terms of Service (effective 23 September 2024) give it a perpetual licence to use what is sent and made “to provide, maintain, promote, and improve the Services”. fal’s own model page marks Creatify Lipsync as run by fal itself, not as a partner API, so your video and audio should stay with fal — which doesn’t train on them — and not reach Creatify; we list Creatify’s terms because they govern the model. Creatify Lab Inc. is in California, United States.

  Its rules for your requests (9)

  • Only lip-sync people whose consent you have, with the rights to use their image and likeness.
  • Don’t impersonate a person or organisation, or misrepresent an affiliation with one.
  • No political campaigning or lobbying.
  • No misinformation, disinformation or conspiracy theories.
  • No pornography or explicit sexual material, and nothing that exploits or abuses children.
  • No threatening, abusive, harassing, defamatory or hateful content.
  • No scams, spam or intellectual property violations.
  • No legal, medical or financial advice without proper authorisation.
  • Don’t use outputs to train AI models or build competing products.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 7 October 2026

• Tripo AI

  Through a licensed API provider

  Not stated

  Models

  Tripo H3.1

  Used for

  Standard-quality 3D models from product photos (through fal)

  Training

  Tripo’s site blocks automated reading, so we couldn’t open its terms; search listings of its help pages say paid users’ content isn’t used for training, but we haven’t confirmed that ourselves. We reach Tripo through fal’s partner API.

  Terms (opens in a new tab)Checked 4 October 2026

• Hyper3D (Deemos)

  Through a licensed API provider

  May train on content

  Models

  Rodin Gen-2.5

  Used for

  Premium 3D models with files for augmented reality (through fal)

  Training

  Deemos’ terms say it may use prompts and outputs “to provide, maintain and improve our services, and conduct and support research”; we reach Rodin through fal’s partner API.

  Its rules for your requests (2)

  • Photos and prompts must not infringe others’ rights, defame anyone, or invade their privacy or publicity rights.
  • No pornographic, racist, homophobic, bullying or offensive prompts, and no one else’s creations or images without permission.

  Terms (opens in a new tab)Checked 4 October 2026

• CassetteAI

  Through a licensed API provider

  Not stated

  Models

  Music Generator, Sound Effects Generator

  Used for

  Background music beds and quick sound effects (through fal)

  Training

  CassetteAI’s terms (13 February 2023) don’t say whether prompts are used for training; we reach it through fal’s partner API.

  Its rules for your requests (2)

  • Music prompts must be your own and must not infringe anyone’s intellectual property; you’re responsible for how you use the music.
  • Don’t resell or sublicense access to the music model.

  Terms (opens in a new tab)Checked 4 October 2026

• Stability AI

  Through a licensed API provider

  No training by default

  Models

  Stable Audio 2.5

  Used for

  Restyling and transforming audio (fal runs it on its own servers)

  Training

  fal runs Stable Audio 2.5 on its own servers (it isn’t one of fal’s partner APIs), so your audio goes to fal — which doesn’t train on it — not to Stability AI.

  Its rules for your requests (8)

  • No child sexual abuse material, sexual exploitation, grooming or trafficking of minors, or suggestive depictions of minors.
  • No non-consensual intimate imagery, illegal pornography, or content about sexual acts or sexual violence.
  • No encouragement of self-harm or harm to others, hateful or violent content about protected groups, extreme gore, or help making illegal weapons.
  • No misinformation, election interference, impersonation without consent or legal right, or pretending AI output was made by a human.
  • No medical or other essential-services advice without review by a qualified professional and disclosure that AI was used.
  • No AI practices banned by EU law, such as manipulative techniques, social scoring, emotion inference at work or school, or biometric categorisation.
  • Nothing that violates intellectual property or privacy rights.
  • Don’t bypass safeguards or account bans, and no malware.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 4 October 2026

• Ideogram

  Through a licensed API provider

  No training by default

  Models

  Ideogram 3.0

  Used for

  Typography, logos and designs with lettering (through fal)

  Training

  Ideogram’s Developer API Agreement: it “shall not use any User Input or User Output to train the Ideogram AI Model”, except content flagged as breaking its usage policies; we reach it through fal’s partner API.

  Attribution

  Ideogram’s API agreement asks apps to say outputs were made by Ideogram’s model (e.g. a “Powered by Ideogram” tagline) where it’s offered, and to link its usage policies. We show “Powered by Ideogram” as text when Ideogram 3.0 is picked; our Acceptable Use Policy links its usage policies.

  Its rules for your requests (7)

  • Nothing harmful to children.
  • No obscene or pornographic content.
  • No defamatory, excessively violent, harassing, threatening, hateful or deceptive content.
  • Don’t upload images of people without their consent, or copyrighted images without the owner’s consent.
  • No political campaigning or lobbying.
  • Don’t mislead anyone by presenting outputs as entirely human-made or as photos of real events.
  • Don’t use inputs or outputs to build competing products.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 4 October 2026

• Recraft

  Through a licensed API provider

  No training by default

  Models

  Recraft V3

  Used for

  Illustration, icons and vector art (through fal)

  Training

  Recraft’s Developer Terms (11 August 2026): “Recraft will not use any Assets created using the API Services to train Recraft AI models”; we reach it through fal’s partner API.

  Its rules for your requests (9)

  • No images that exploit or abuse children.
  • No non-consensual intimate imagery or sexual deepfakes of real people.
  • Nothing that infringes intellectual property, privacy or publicity rights.
  • Nothing threatening, abusive, harassing, defamatory, deceptive, fraudulent, obscene or promoting self-harm.
  • No false information meant to harm others, impersonation, or sharing of identifying personal information without authorization.
  • Nothing supporting terrorist organisations or promoting violence against protected groups.
  • When prompting in the style of a brand, creator or real person, you must have the rights and consents to do so.
  • Don’t use outputs to train AI models or build competing products.
  • Don’t remove metadata, watermarks or other markings Recraft adds.

  Terms (opens in a new tab)Usage policy (opens in a new tab)Checked 4 October 2026

• Bria AI

  Through a licensed API provider

  No training by default

  Models

  RMBG 2.0

  Used for

  Background removal in the Background Remover, Product Ads, the Product Launch Kit and the Design Studio (through fal’s partner API)

  Training

  Bria’s background-removal terms (September 2026) let it use content only “for the performance and enhancement of the Products for Customer’s use only” and say it “shall not be used to train generative models” without written consent; Bria is based in Tel Aviv, Israel, its terms incorporate its data processing addendum, and fal marks this model Enterprise Ready, so fal’s own addendum covers it.

  Its rules for your requests (8)

  • Only remove backgrounds from images you have the rights to — nothing that infringes copyright or other people’s privacy and publicity rights.
  • Nothing deceptive under consumer protection laws, and nothing that impersonates another person.
  • Nothing that promotes illegal drugs, illegal gambling, illegal arms trafficking or export-control breaches.
  • Nothing unlawful, defamatory, threatening, pornographic, harassing, hateful or racially or ethnically offensive.
  • No content implying endorsement by political parties or opinion-based movements.
  • Nothing malicious or fraudulent.
  • No use in EU AI Act high-risk systems or other environments that need fail-safe performance.
  • Don’t use RMBG 2.0 outputs to train AI models (for example by distillation).

  Terms (opens in a new tab)Checked 4 October 2026

Their rules are summarised in our AI & Acceptable Use Policy, and how your content is handled is explained in the Privacy Policy.

• Powered by Runway. Models marked with the Runway badge run on Runway’s developer platform. We show that badge wherever you use one of them, as Runway’s terms require.
• Licensed API providers. Other models run through licensed API providers, either on the provider’s own systems or on the developer’s.
• Directly from the developer. OpenAI’s models are used through OpenAI’s own API.
• The developers’ usage policies apply to requests you send to their models. Which providers receive your data, and where they are, is listed on the Third-Party Processors page.

AI models are updated and retired often. When a developer retires a model, we remove it from QikTools and, where needed, move the tool to another model and tell you. We don’t advertise models we can’t run. Some tools use a second model for part of a job, for example an editing model when you add a reference picture, and the price shown always covers the whole job.

Kling, Seedance, Seedream, Veo, Gemini, Nano Banana, Lyria, GPT, Runway, Gen-4.5, GWM-1, Wan, Hailuo, MiniMax, ElevenLabs, FLUX, Ideogram, Recraft and other product or company names are trademarks of their respective owners. QikTools uses them only to identify third-party AI models available through our service. QikTools is an independent service and is not affiliated with, sponsored by or endorsed by these companies. Model availability, features and pricing may change.

If you own one of these marks and think we have used it wrongly, write to hello@qiktools.com and we will fix it. Our own marks are covered in the Terms of Service.