Free AI for everyone

The best FREE AI

The free AI tools that are genuinely worth it, tested and hand-picked by us, with no inflated promises. We've split them into six areas: assistants, images, code, video, audio and the models you install on your own computer. Pick the one you need below.

Tap a button and that AI opens, ready to use. “Info” shows the real limits of the free plan and where your data ends up: there is no such thing as truly “unlimited”, and we say so.

Assistants

The all-rounders: writing, translating, reasoning and searching the web.

10 toolsyou are here

Images

Generate images from text: realistic photos, art, logos.

9 toolsGo→

Code

Programming: assistants in your editor and online builders.

8 toolsGo→

Video

Create video clips from text or from a photo.

3 toolsGo→

Audio

Music and synthetic voices, starting from text.

5 toolsGo→

Local AI

Models to install on your PC: free, offline, private.

GuideGo→

Assistants · The all-rounders: writing, translating, reasoning and searching the web.

ChatGPT
GPT-5.6 Luna
ChatGPT · GPT-5.6 Luna
OpenAI
The chatbot that kicked off the AI race. On the free plan you get GPT-5.6 Luna, with a Think button to make it reason for longer: it writes, reasons, generates images and reads the files you upload.
Performance 4.5 · Free plan 5 → 5 stars out of 5
Effectively unlimited
Text chats have had no cap since 6 August 2026, with an abuse safeguard. ⚠️ Files, images, voice and data analysis have separate limits, and OpenAI does not publish the numbers. Since 18 August 2026 ads may appear on the free plan, in Europe too.
Data in USA
OpenAI servers (USA). By default your chats can be used for training: you can switch it off in the settings.
GPT-5.6 Luna · freeGPT-6 Luna (desktop app only) · freeGPT-5.6 Sol · paidGPT-6 Pro · paid
Open ChatGPT
DeepSeek
V4 Pro
DeepSeek · V4 Pro
DeepSeek
A powerful Chinese model, free and with no subscription: in the app you choose between Instant mode and Expert mode (DeepSeek V4 Pro), with a switch to make it reason.
Performance 4 · Free plan 4.5 → 4.5 stars out of 5
Effectively unlimited
The chat has no subscription and no stated message caps. There is no official page that guarantees it, though: it is a fact, not a promise. ⚠️ What is free is the chat: the developer API is pay-as-you-go.
Data in China
Warning: data is collected and stored in China and your inputs may train the models (you have a right to object). Better not to enter sensitive or business data.
DeepSeek V4 Pro (Expert) · freeDeepSeek Flash (Instant) · free
Open DeepSeek
Qwen Studio
Qwen3.8
Qwen Studio · Qwen3.8
Alibaba
Alibaba’s AI, now Qwen Studio: multilingual, multimodal and generous. On the free plan you get Qwen3.8, and the family’s models are open (Apache licence), downloadable even in the largest version.
Performance 4 · Free plan 4 → 4 stars out of 5
Effectively unlimited
The chat has no cap communicated, and it holds up under heavy use. ⚠️ Free access to Qwen Code for developers, on the other hand, closed on 15 April 2026.
Data in China
Servers in China (Alibaba). There is no clear statement on training: as with DeepSeek, avoid confidential data.
Qwen3.8 · freeQwen3.8 (open weights, local) · free
Open Qwen Studio
Gemini
3.6 Flash
Gemini · 3.6 Flash
Google
On the free plan you get Gemini 3.6 Flash, Google’s fast model: multimodal (it understands text, images, audio and code) and connected to Search, Android and Workspace. Now and then it also lets you use 3.1 Pro, in variable doses.
Performance 4.5 · Free plan 3 → 4 stars out of 5
Capped
⚠️ Google does not publish the numbers. Since May 2026 the limit is based on the compute you use and refills every 5 hours up to a weekly cap: a complex question weighs more than a simple one. Deep Research is included, but may be unavailable at busy times. There is no video; 3.7 and 3.8 Flash are for paying users only.
Data in USA
Google servers (USA). With "Activity" on, your chats help improve the models and some are read by human reviewers, who can keep them for up to 3 years: turn "Activity" off to stop it.
Gemini 3.6 Flash · freeGemini 3.1 Pro (in doses) · freeGemini 3.7 / 3.8 Flash · paidDeep Think · paid
Open Gemini
Meta AI
Muse Spark
Meta AI · Muse Spark
Meta
Meta’s assistant, inside WhatsApp, Instagram, Facebook and Messenger. Since April 2026 it runs on Muse Spark, Meta’s new model and one of the best rated in the rankings. ⚠️ Unlike the old Llama models, it is not open source.
Performance 4 · Free plan 3 → 3.5 stars out of 5
Capped
Free inside Meta’s apps, and Meta does not state any number. ⚠️ The real point for European users: it is not confirmed that Muse Spark is already live in Italy (Meta says availability varies by country), and some features are missing in the EU.
Data in USA
Meta servers (USA). In the EU Meta uses public content and your conversations with Meta AI for training, with a form to object; private messages are excluded.
Muse Spark · free
Open Meta AI
Perplexity
automatic choice
Perplexity · automatic choice
Perplexity
More an answer engine than a chatbot: it searches the web in real time and answers always citing its sources. On the free plan it picks the model itself, the one that fits the question.
Performance 4 · Free plan 3 → 3.5 stars out of 5
Capped
Basic searches are practically unlimited. The ones that matter are few: 3 Pro Searches a day and 1 Research a month, plus a file limit. The best models (Sonar 2, GPT, Claude) are in the Pro and Max plans.
Data in USA
Perplexity servers (USA). Data retention for training is on by default on the free plan too: switch it off in Preferences.
Automatic choice ("Best") · freeSonar 2, GPT-5.6, Claude Sonnet 5 (Pro) · paid
Open Perplexity
Claude
Sonnet 5.5
Claude · Sonnet 5.5
Anthropic
Anthropic’s AI (the same family that powers this site). Since 28 September 2026 the free plan runs Claude Sonnet 5.5, one of the most capable models in the world today: it excels at long texts, careful writing, reasoning and code.
Performance 5 · Free plan 2 → 3.5 stars out of 5
Capped
⚠️ Anthropic does not publish the number: it only says the free plan has 5-hour sessions and that Pro gives at least 5 times as much. The "10-20 messages" or "15-40" you read around are not official and do not agree. The top model, Opus 5.5, is subscription-only.
Data in USA
Anthropic servers (USA). Since 2025 you choose whether your chats are used for training: if you agree they are kept for 5 years, if you refuse for 30 days. Check your choice in the privacy settings.
Claude Sonnet 5.5 · freeClaude Haiku 4.5 · freeClaude Opus 5.5 · paid
Open Claude
Vibe (Le Chat)
Medium 3.5
Vibe (Le Chat) · Medium 3.5
Mistral
The chatbot from France’s Mistral, renamed Vibe in August 2026 (same address and account as Le Chat): the main European alternative, with data processed in the EU. It runs Mistral Medium 3.5, which is fast.
Performance 3 · Free plan 3 → 3 stars out of 5
Capped
⚠️ Mistral does not state the numbers: its pricing page only says "limited messages and web searches". In writing: image generation and $10 a month in API credits. The Pro plan confirms the cap exists, sold as 6× the messages and 40× the images of the free plan.
Data in EU
French company, data processed in the EU. ⚠️ But on the free plan your conversations train the models by default: switch it off in the settings ("Allow your interactions to be used to train our models").
Mistral Medium 3.5 · freeAgents and long tasks (Pro) · paid
Open Vibe (Le Chat)
Grok
4.5
Grok · 4.5
xAI
xAI’s AI, built into the X social network. On the free plan you get Grok 4.5, including voice, with one distinctive trait: access to real-time news and a blunt, unflattering tone.
Performance 4 · Free plan 2 → 3 stars out of 5
Capped
Chat and voice have free-tier limits xAI does not publish: you see the count inside the app. ⚠️ Images and video (Grok Imagine) have been paid-only since March 2026: you need a SuperGrok plan.
Data in USA
xAI servers (USA). Conversations train the models by default: switch off "Improve the model" or use Private Chat.
Grok 4.5 (chat and voice) · freeGrok Imagine (images and video) · paidSuperGrok and Heavy · paid
Open Grok
Copilot
GPT models
Copilot · GPT models
Microsoft
Microsoft’s assistant, built on OpenAI models and woven into Windows, Edge and Office. It is already on your PC, ready to summarise, write and generate images with nothing to install.
Performance 3.5 · Free plan 2 → 3 stars out of 5
Capped
⚠️ With the new unified app (rolled out from August to October 2026), free users hit a daily chat cap, and Microsoft does not publish the number: users report anywhere from 2 to 10 messages. Microsoft does not say which GPT model the free version uses.
Data in USA
Microsoft servers (USA). If you sign in, your conversations may be used for training: you can opt out in the privacy settings. Without signing in they are not used.
Copilot (free, capped) · freeMicrosoft 365 Personal / Premium · paid
Open Copilot

The stars are the average of two scores from 1 to 5, with the same rule for every tool: performance (how good it is, from independent tests and rankings) and free plan (how much it lets you do without paying, from the limits the makers publish).

Images · Generate images from text: realistic photos, art, logos.

Microsoft Designer
MAI-Image
Microsoft Designer · MAI-Image
Microsoft
Microsoft’s generator, free with a Microsoft account (also available as Bing Image Creator). It now uses Microsoft’s own MAI-Image models instead of the old DALL·E 3: type and get it, in the browser.
Performance 4 · Free plan 3.5 → 4 stars out of 5
Capped
Microsoft only says it is free and that a subscription may be needed "to create more often": no number. Bing Image Creator gives 10 fast creations, then unlimited slower ones. ⚠️ Designer’s FAQ speaks of personal, non-commercial use. No visible watermark.
Data in USA
Microsoft servers (USA). Fairly strict content filters; images carry invisible provenance metadata.
MAI-Image 2.5 Flash · freeMore fast creations (subscription) · paid
Open Microsoft Designer
Flux (local)
FLUX.2 klein
Flux (local) · FLUX.2 klein
Black Forest Labs
The open models from Black Forest Labs (Germany, the original authors of Stable Diffusion). FLUX.2 [klein] 4B is free, commercial use included: you download it and run it on your own computer, with no account and without your images leaving home.
Performance 3.5 · Free plan 3.5 → 3.5 stars out of 5
Effectively unlimited
On your own computer there is no cap and no watermark, but you need a graphics card with about 13 GB of memory (RTX 3090 or 4070 and up) and some know-how. It is the freest free route, not the easiest. The Pro versions on online services are paid.
Data in EU
Black Forest Labs is German. Run locally, your images never leave your computer.
FLUX.2 klein 4B · freeFLUX.1 schnell · freeFLUX Pro (online) · paid
Open Flux (local)
Leonardo.ai
Phoenix · Lucid
Leonardo.ai · Phoenix · Lucid
Leonardo (Canva)
A complete, easy platform, now part of Canva: you generate with Leonardo’s models and get retouching and upscaling tools. Every day you receive free tokens that refill on their own.
Performance 3.5 · Free plan 3.5 → 3.5 stars out of 5
Capped
150 tokens a day, which do not accumulate. ⚠️ On the free plan your creations are public and remain Leonardo’s property, with a licence for you to use them: for confidential work you need a paid plan. Leonardo does not specify which models are included.
Data in USA
Australian company (Canva group), international servers. Avoid images with personal data.
Phoenix · freeLucid · freeUnlimited slow generations · paid
Open Leonardo.ai
Gemini (images)
Nano Banana 2
Gemini (images) · Nano Banana 2
Google
The image generator inside Gemini: describe what you want and you get it, with nothing to install. On the free plan you get Nano Banana 2; the more powerful Nano Banana Pro is for subscribers.
Performance 4 · Free plan 3 → 3.5 stars out of 5
Capped
⚠️ Google does not publish a number: the limit is based on compute, refills every 5 hours up to a weekly cap, and images weigh more than text. Since 14 August 2026 the visible watermark can be removed in the settings ("Media watermark"); the invisible one (SynthID) always stays.
Data in USA
Google servers (USA). With "Activity" on, your content helps improve the models: switch it off in the settings.
Nano Banana 2 · freeNano Banana Pro · paid
Open Gemini (images)
Krea
Krea 2
Krea · Krea 2
Krea
A creative tool with real-time preview: it brings its own Krea 2 models and several open models (Flux, Z Image) together in one polished interface.
Performance 4 · Free plan 3 → 3.5 stars out of 5
Capped
100 compute units a day, which refill: for example about 50 images with Krea 2 Turbo, 11 with Krea 2 Medium or 5 with Krea 2 Large. ⚠️ The free plan has no commercial licence (it comes with Basic), and the best outside models (Nano Banana, GPT Image) are excluded.
Data in USA
US servers. Fine for creative images, avoid sensitive data.
Krea 2 (Turbo, Medium, Large) · freeFlux.1 Dev, Z Image · freeNano Banana, GPT Image · paid
Open Krea
Adobe Firefly
Nano Banana + Firefly
Adobe Firefly · Nano Banana + Firefly
Adobe
Adobe’s AI, inside Photoshop and Express. On the free plan you get daily generations on a selection of models, including Google’s Nano Banana and the Firefly models, for video and audio too.
Performance 4 · Free plan 2.5 → 3.5 stars out of 5
Capped
The free plan offers daily generations that reset every day, on a curated set of models: Adobe does not publish the number (and its FAQ still contains an old paragraph about monthly credits). ⚠️ The copyright safety applies only to the Firefly models: for partner models such as Nano Banana, Adobe does not give the same guarantee.
Data in USA
Adobe (USA). Your content is not used to train the models without consent.
Nano Banana · freeFirefly models (selection) · freeGenerative credits · paid
Open Adobe Firefly
ChatGPT (images)
Images 2.5
ChatGPT (images) · Images 2.5
OpenAI
Generating images by talking: ask, look and fix it in plain words ("darker", "remove the chair"). Since 8 September 2026 free users get ChatGPT Images 2.5 too, currently first in the human-vote rankings, and excellent at writing text inside images.
Performance 5 · Free plan 2 → 3.5 stars out of 5
Capped
⚠️ Images are the exception to ChatGPT’s "no cap": text is unlimited, images are not. OpenAI does not publish the number; users report 2-3 images a day over a 24-hour window. Images "with thinking" are paid-only.
Data in USA
OpenAI servers (USA). By default your chats can be used for training: you can switch it off in the settings.
ChatGPT Images 2.5 · freeImages with thinking · paid
Open ChatGPT (images)
Reve
Reve 2
Reve · Reve 2
Reve AI
A young generator that is already among the best in the human-vote rankings (sixth in September 2026, above Nano Banana 2). A simple interface: type, get, fix.
Performance 4.5 · Free plan 2 → 3.5 stars out of 5
Capped
⚠️ The free plan exists, but Reve does not publish the number of images: it only says the paid plans give 5 and 100 times as many. Usage rights and watermarks on the free plan are not clearly stated: check before using them for work.
Data in USA
Reve AI (USA).
Reve 2 · freeMore images (Lite, Pro) · paid
Open Reve
Ideogram
slow queue
Ideogram · slow queue
Ideogram
The image AI best at writing correct text inside graphics: logos, posters, lettering. Since June 2026 its Ideogram 4.0 is also an open-weights model you can run locally.
Performance 4 · Free plan 2 → 3 stars out of 5
Capped
⚠️ Weekly credits, and Ideogram does not publish the number ("may vary"); you need to sign in with Google, Apple or Microsoft. Slow queue only, one generation at a time; AI editing is paid. On the free plan images are public; commercial use is allowed. Ideogram says nothing about watermarks.
Data in USA
US servers. On the free plan every image goes to the public gallery.
Ideogram (slow queue) · freeIdeogram 4.0 (open weights, local) · freeFast credits and AI editing · paid
Open Ideogram

Code · Programming: assistants in your editor and online builders.

Local Code Assistants

You install them in your editor (VS Code, JetBrains) or as a standalone app: they complete the code as you write and work on your project, on your computer. For people who really code.

Google Antigravity
Gemini Flash
Google Antigravity · Gemini Flash
Google
Google’s environment for coding with agents, which replaced Gemini Code Assist for individuals: editor, command line and extensions for VS Code and JetBrains. On the free plan you get the Gemini Flash models.
Performance 4 · Free plan 3.5 → 4 stars out of 5
Capped
Free, with unlimited completions and commands and a weekly agent limit that Google does not publish. ⚠️ Google’s own sources disagree on the included models: the pricing page also lists Gemini 3.1 Pro and Claude, the documentation only the Gemini Flash models.
Data in USA
Google (USA). Your interactions help improve its products and may be read by people: you can switch it off in the settings.
Gemini 3.8 / 3.7 / 3.6 Flash · freeHigher quotas (Google AI Pro, Ultra) · paid
Open Google Antigravity
GitHub Copilot
Auto
GitHub Copilot · Auto
GitHub (Microsoft)
GitHub’s assistant, the de facto standard, with a free plan. It completes code as you type, answers in chat and has an agent mode; it installs in VS Code, Visual Studio and JetBrains.
Performance 4 · Free plan 2.5 → 3.5 stars out of 5
Capped
2,000 completions a month. For chat, since 1 June 2026 the Free plan has an allowance of AI credits that GitHub does not quantify. ⚠️ Since 24 June 2026 free users cannot choose the model: the system picks it ("Auto").
Data in USA
⚠️ Since 24 April 2026, on the Free, Pro and Pro+ plans your prompts and suggestions are used to train the models unless you switch it off in the Copilot settings. GitHub/Microsoft (USA).
Auto (model chosen by GitHub) · freeModel choice · paid
Open GitHub Copilot
Devin Desktop
free models
Devin Desktop · free models
Cognition
Cognition’s editor: it is the old Windsurf, renamed Devin Desktop in June 2026. You install it and code with AI help; the autonomous cloud agent, Devin proper, is paid.
Performance 4 · Free plan 2.5 → 3.5 stars out of 5
Capped
The Free plan has unlimited completions and inline edits and a "light quota" for agents, which renews daily and weekly: Cognition does not publish the numbers. The free models do not count against the quota. Cloud agents start from the Pro plan.
Data in USA
⚠️ By default Cognition may use your data to train its models, and on the free plan you cannot opt out: only paying users can.
Free models · freeDevin Cloud (agents) · paid
Open Devin Desktop
Cursor
Composer
Cursor · Composer
SpaceX (formerly Anysphere)
The AI-first editor: a souped-up VS Code where the AI knows the whole project and works in agent mode. Since August 2026 it has been part of SpaceX. The free plan is called Hobby.
Performance 4.5 · Free plan 2 → 3.5 stars out of 5
Capped
The Hobby plan is free, with no card, "limited agent requests" and access to Composer. ⚠️ Cursor does not publish any number. For daily use you need the Pro plan.
Data in USA
With privacy mode off, Cursor may use your code to train its models: turn it on. Cursor does not say whether it is on by default for free users.
Composer · freeFrontier models · paid
Open Cursor

Web Code Assistants

You use them online in the browser, with nothing to install: you describe what you want and they build the app. Perfect for getting started fast, for prototypes and for anyone without a development environment.

Google AI Studio
Gemini 3.8 Flash
Google AI Studio · Gemini 3.8 Flash
Google
Google’s online lab for building with Gemini: try the models, generate code and prototype apps, free, from the browser.
Performance 4 · Free plan 4 → 4 stars out of 5
Capped
Google says AI Studio is free in all regions; the quotas are not published, you only see them in your dashboard. The Gemini Flash models are free (3.8, 3.7, 3.6, 3.5); Gemini 3.1 Pro is not.
Data in USA
⚠️ On the free plan your content is used to improve Google’s products: it matters if you paste in a client’s code.
Gemini 3.8 Flash · freeGemini 3.1 Pro · paid
Open Google AI Studio
bolt.new
Standard agent
bolt.new · Standard agent
StackBlitz
From prompt to app, in the browser: describe the application and bolt builds it, previews it and lets you download or publish it.
Performance 4 · Free plan 3.5 → 4 stars out of 5
Capped
300,000 tokens a day and 1 million a month, public or private projects, Bolt-branded hosting, uploads up to 10 MB. Enough for prototypes and small web apps; on big projects the tokens run out fast.
Data in USA
StackBlitz (USA). From 7 October 2026 it uses code and chats to train models, but European accounts are excluded.
Standard agent · freeMax agent · paid
Open bolt.new
Replit
Free Mode
Replit · Free Mode
Replit
A development environment in the browser with AI: describe what you want and the agent creates, edits and runs the app, with nothing to install.
Performance 4 · Free plan 2 → 3 stars out of 5
Capped
The Starter plan gives daily agent credits, up to a monthly cap: ⚠️ Replit does not publish the numbers. Only the "Lite" version of the agent and a single published app, which expires after 30 days, branded "Made with Replit".
Data in USA
Replit (USA). Its privacy policy says nothing about training on your projects.
Free Mode (model chosen by Replit) · freeFull agent · paid
Open Replit
v0
v0
v0 · v0
Vercel
Vercel’s AI for interfaces: describe a screen and v0 generates clean React/Next code, ready to use.
Performance 4 · Free plan 2 → 3 stars out of 5
Capped
7 messages a day. Free credits "follow the offer that applies to your account", with no fixed number. For continuous work you need the Plus plan ($30 a month).
Data in USA
⚠️ On the free plan you cannot exclude your data from training: the opt-out exists only from the Business plans. Vercel (USA).
v0 (free) · freeAll models (Plus) · paid
Open v0

Video · Create video clips from text or from a photo.

Google Flow
Veo 3.1 Lite
Google Flow · Veo 3.1 Lite
Google
Google’s video studio, with the Veo 3.1 and Gemini Omni models, among the best in the world today (Gemini Omni is first in the human-vote rankings). ⚠️ In the Gemini app, free users no longer get video: the free route is Flow.
Performance 5 · Free plan 3.5 → 4.5 stars out of 5
Capped
50 credits a day, even without a subscription, which do not accumulate: a clip with Veo 3.1 Lite costs 10, with Gemini Omni Flash 7 to 15, so about 4-5 clips a day. The invisible SynthID watermark is always there; the visible one can be turned off. ⚠️ Flow’s "Get started" page still asks for a subscription while the credits page says it does not: Google is not consistent.
Data in USA
Google (USA). You must be 18; Google does not claim ownership of your videos.
Veo 3.1 Lite · freeGemini Omni Flash · freeVeo 3.1 Quality · paid
Open Google Flow
Dreamina
Seedance 2.0
Dreamina · Seedance 2.0
CapCut (ByteDance)
CapCut’s creative platform (ByteDance), with the Seedance 2.0 video model, fifth in the human-vote rankings. Video from text or from an image, in the browser.
Performance 4.5 · Free plan 3 → 4 stars out of 5
Capped
120 credits a day according to the official FAQ (some third-party sources say 225). ⚠️ On the free plan videos come with a watermark, and for commercial use Dreamina refers you to its terms.
Data in Asia and USA
ByteDance stores data on servers in Malaysia, Singapore and the United States, and says it also uses it to train its models. Avoid faces and confidential material.
Seedance 2.0 · freeWatermark-free video · paid
Open Dreamina
Kling
Kling 3.0
Kling · Kling 3.0
Kuaishou
Kuaishou’s video generator, one of the most used: clips from text or from an image, with smooth motion and audio too. Every day you receive free credits.
Performance 4 · Free plan 3 → 3.5 stars out of 5
Capped
Free credits every day, which do not accumulate: ⚠️ Kling does not publish the number (third-party sources say 66). At Kling 3.0’s stated cost that is about one 5-second clip with audio a day, two without. On the free plan videos come with a watermark, and the Terms forbid commercial use without written permission.
Data in Singapore
Kling AI Pte. Ltd. (Kuaishou group) stores data in Singapore. Avoid faces or confidential material.
Kling 3.0 · freeKling 4.0 · paid
Open Kling

Audio · Music and synthetic voices, starting from text.

Google AI Studio (voice)
Voice · Gemini 3.8 Flash TTS
Google AI Studio (voice) · Gemini 3.8 Flash TTS
Google
Gemini’s speech synthesis inside AI Studio: type the text, pick a voice and listen to it in 130 languages, Italian included. Fourth in the human-vote ranking.
Performance 4 · Free plan 3.5 → 4 stars out of 5
Capped
On the free tier it is free of charge; the quotas are not published, you see them in your dashboard. ⚠️ Fine for personal use: anyone offering an app to European users must switch to the paid plan.
Data in USA
⚠️ On the free tier your texts may be read by human reviewers and used to improve Google’s products.
Gemini 3.8 Flash TTS · free
Open Google AI Studio (voice)
ElevenLabs
Voice · Eleven v4
ElevenLabs · Eleven v4
ElevenLabs
The standard for synthetic voices: it reads your texts with realistic voices in many languages, Italian included. The new Eleven v4 is first in the human-vote ranking, and it is on the free plan too.
Performance 5 · Free plan 2.5 → 4 stars out of 5
Capped
10,000 credits a month (about 10 minutes of voice). Files can be downloaded as MP3 or WAV. ⚠️ On the free plan no commercial use and attribution is required ("elevenlabs.io" in the title); voice cloning is paid.
Data in USA
ElevenLabs (USA). Cloning a voice requires the person’s consent.
Eleven v4 · freeEleven v3, Multilingual v2 · freeVoice cloning · paid
Open ElevenLabs
Gemini (music)
Music · Lyria 3.5
Gemini (music) · Lyria 3.5
Google
Music generation inside Gemini, with Google’s Lyria 3.5 model: since 4 September 2026 it is available to everyone, for short or long tracks, sung or instrumental.
Performance 4 · Free plan 3 → 3.5 stars out of 5
Capped
It is on the free plan too and tracks can be downloaded, with the invisible SynthID mark. ⚠️ Google does not publish the limits nor clearly say whether you can use them for work; you must be 18.
Data in USA
Google servers (USA). Your account’s "Activity" settings apply.
Lyria 3.5 · free
Open Gemini (music)
Stable Audio
Music · Stable Audio 3.0
Stable Audio · Stable Audio 3.0
Stability AI
Stability AI’s music generator: instrumental tracks and sound effects from text, useful as background music. The model is trained only on licensed material.
Performance 3.5 · Free plan 2.5 → 3 stars out of 5
Capped
50 credits a month, about 12 full tracks, for personal, non-commercial use only. The commercial licence starts with the Solo plan ($12 a month). Stability does not say whether free users can download.
Data in UK
Stability AI (United Kingdom). Built for music and effects, not spoken voice.
Stable Audio 3.0 · freeCommercial licence · paid
Open Stable Audio
Suno
Music · v6-mini
Suno · v6-mini
Suno
The most popular tool for making music from text: describe the genre and theme and you get a complete song with vocals. The free v6-mini model is second in the rankings, ahead of every competitor.
Performance 4.5 · Free plan 1.5 → 3 stars out of 5
Capped
50 credits a day. ⛔ But since 3 September 2026 the free plan no longer lets you download songs: those who signed up earlier get at most 7 trial downloads, newcomers get them "occasionally". No commercial use. You listen and share, but you cannot take the file home.
Data in USA
Suno (USA). Songs start out visible only to people with the link; they cannot be used commercially.
Suno v6-mini · freeSuno v6 and v6-wild · paid
Open Suno

AccademIA

AI for Schools and Universities

Studying or teaching? AI tools for schools and universities, checked one by one: real prices, minimum age, data.

Open AccademIA

Want more of this?

In the members area, free and with no password, you'll find AI news and GitHub Best: trending open source AI repositories on GitHub, with skills, plugins, MCP connectors and models you can run on your own computer.

Enter the members area →

This is just the beginning.

PlayMind Studio helps companies and professionals use AI for real: consulting, training and tailor-made solutions.

See what we do →

Free plans change all the time: we keep this list up to date. If you find a mistake or something is missing, write to us.