Rewind.ai
Rewind.ai is a browser-based platform bundling 400+ AI tools and 400+ models — chat, image, video, voice, music, writing, code, transcription, translation — behind one token balance, with a permanent free tier that works without an account.
What is Rewind.ai?
Rewind.ai is a browser-based platform that gathers 400+ AI tools across 18 announced categories and 400+ models behind a single account and a single token balance. The catalogue is split into twenty sections: AI Chat, AI Search, AI Image (72 tools), AI Video (28), AI Voice (14), AI Music (28), AI Writer (41), AI Code (31), AI Design (28), Transcription (12), AI Translate (7), Text Tools (21), PDF Tools (20), Study Tools (18), YouTube Tools (20), Business Tools (24), Social Media (22), Health & Lifestyle (16), Website Builder (14), plus a model directory listing 495 entries.
One clarification comes first. This site is not the original Rewind AI — the Mac app and Pendant that later became Limitless, was bought by Meta and shut down on 19 December 2025. The current operator says so itself: it declares no affiliation with Rewind AI, Limitless or Meta, and describes the domain as hosting, since 2026, a different product built by a different team. Nothing from the earlier company — its founders, its funding, its recording hardware — carries over.
Most tools run without an account. Signing up is free, triples the daily quota, keeps generation history and syncs settings across devices. One balance covers every tool, and the token cost of a request is displayed before you send it.
The economics are unusually well documented. Open-weight models — Qwen 2.5 and 3.5, FLUX, Kokoro, Whisper, ACE-Step, Wan — run on GPUs the operator owns, which is how a permanent free tier is presented as sustainable; external models such as GPT-4o, Claude, Gemini, Llama, DeepSeek, Grok, Mistral and Phi are proxied and billed at provider cost, with no claimed markup. The stack is published too: a Next.js front end, a Django 6.x back end, a FastAPI inference service on NVIDIA A100 GPUs hosted at Vultr, PostgreSQL, Redis and Nginx.
The home page itself is configurable — reorder tabs, remove them, pin a tool as the main tab — and those settings follow the user from one device to another. On-site counters claim 400+ tools, 400+ models, 2K creators and 25.0M tokens consumed. No team, no legal entity and no address are named anywhere on the site.
What it does
- Chat with 400+ models from one account, from self-hosted Qwen to proxied GPT-4o, Claude, Gemini, Llama, DeepSeek, Grok and Mistral
- Generate and retouch images with FLUX: editing, upscaling, background and object removal, face swap, portrait generation
- Create video, subtitles and dubbing with Seedance and Wan, including shorts, reels and animation
- Synthesise speech with Kokoro, transcribe live audio, meetings, podcasts and calls with Whisper, change voices and generate music
- Write, rewrite, proofread and translate across 100+ languages, and check text for AI-generated content
- Work on documents: summarise or chat with a PDF, run OCR, convert PDF files to Word or Excel
- Generate, explain and review code in Python, JavaScript, HTML, CSS, SQL or regex — or call every tool through one OpenAI-compatible REST API key
When to use Rewind.ai / When not to
A quick filter to help you decide if Rewind.ai is the right fit.
When to use Rewind.ai
- Generalists who want to sample several families of AI — chat, image, video, voice, writing, code — without paying for a separate subscription per tool
- Budget-conscious individuals and small teams: a permanent free tier of 5,000 tokens a day, no credit card, and paid packs starting at USD 5
- Developers looking for a single OpenAI-compatible REST endpoint that covers chat, images, text-to-speech, speech-to-text, video and translation
- Freelancers and creators who need commercial rights on what they produce: outputs belong to the user, for personal or commercial work
- Small teams on the Business plan at USD 49 a month, with unlimited seats drawing on one shared token pool and per-member spending caps
When not to use Rewind.ai
- Anyone who needs a native mobile or desktop application: the platform runs in a browser only, with nothing to install
- Buyers with a formal vendor due-diligence process: no legal entity, postal address, named officer, hosting country, subprocessor list or SOC 2 / ISO certification is published
- Teams that require an explicit GDPR compliance commitment or an Article 27 EU representative — the site says it operates on GDPR principles but never claims compliance outright
- Professionals seeking medical, legal or financial guidance: the terms exclude reliance on outputs as professional advice and place verification on the user
- Heavy single-purpose users running one model at high volume, for whom per-token billing can cost more than a dedicated subscription
How to use Rewind.ai
A typical end-to-end flow, from setup to results.
- Open any tool from the catalogue or from the tabs on the home page and run a request straight away, with no account
- Expect anonymous quotas of around 20 chat messages, 5 image generations and 10 writing tasks per day, reset every 24 hours
- Create a free account from an email address alone — no password at sign-up, a link is sent afterwards — for 10,000 bonus tokens, a tripled daily quota and saved history
- Watch the token cost shown before each request: one single balance is debited whichever tool you use
- Compare outputs in the playground by picking a model and adjusting temperature and maximum tokens
- Add a token pack or a subscription through Stripe when you need more; the 5,000 free tokens a day stay active on every plan, paid ones included
- For programmatic use, generate an sk-rewind- API key once tokens are on the account
- Point an OpenAI-compatible client at the Rewind.ai API base URL and call the v1 endpoints: chat/completions, images/generations, videos/generate-async, tts and stt
- For asynchronous video generation, keep the returned job id and poll the jobs endpoint until the render is ready
- Connect the rest: Slack through webhooks, WordPress through a shortcode, Zapier through its Webhooks action; on team plans, share one token pool across unlimited seats with daily, weekly or monthly caps per member
Pros & Cons
Pros
- A permanent free tier of 5,000 tokens a day, usable without an account or a credit card, and kept on every paid plan
- One token balance for 400+ tools instead of one subscription per tool, with the cost of each request displayed before you send it
- External models are announced as billed at provider cost, with no markup added
- Unusual technical transparency: the stack (Next.js, Django 6.x, FastAPI, A100 GPUs at Vultr) is published, and the site itself explains that it is not the original Rewind AI
- Self-hosted models carry permissive Apache 2.0 or MIT licences with weights published on HuggingFace and GitHub, and outputs may be used commercially
- An explicit no-training commitment on user content, covering free, paid and enterprise accounts, with 2FA available everywhere
- OpenAI-compatible API — switching means changing the base URL and the key — and 100 languages listed by name with their ISO codes
Cons
- No legal entity, postal address or named officer is published anywhere, which would make legal recourse difficult
- A single contact channel, support@rewind.ai, obfuscated by Cloudflare in the page source, with no dedicated legal or GDPR address
- No explicit GDPR compliance statement, no Article 27 EU representative, no declared hosting country or region, no published subprocessor list and no certification on display
- No mobile or desktop application; the Chrome extension is still announced as coming to the Chrome Web Store and the Firefox version is only planned
- The site contradicts itself on scope: 400+ tools on the home page and the catalogue, 200+ on the enterprise and integrations pages
- The domain was re-registered on 1 May 2026, so the current platform is very recent and has no verifiable track record
- No quantified retention policy — data is kept “only as long as necessary” — and lasting confusion with the original Rewind AI remains likely despite the clarification page
Pricing & Plans
A free plan is available and permanent: 5,000 tokens per day with no credit card required, plus 10,000 bonus tokens on sign-up, and the daily free allowance remains active on every paid plan. The lowest paid entry point is the Basic plan at USD 5.00 per month for 200,000 tokens, followed by Pro at USD 19 per month and Business at USD 49 per month; annual billing is advertised at 20% off. One-off token packs without a subscription start at the same USD 5.00 for 200,000 tokens and carry no expiry date. Enterprise pricing is quoted on request, with token rates decreasing by volume. Payments are processed by Stripe, and subscriptions can be cancelled at any time, tokens remaining available until the end of the paid period.
- Free — 5
- 000 tokens per day
- 10
- 000 tokens on sign-up
- access to every tool and free use of the self-hosted models
- 200
- 000 tokens per month
- everything in Free
- generation history and priority processing (roughly 100 images
- 10 video generations or 1
- 333 chat messages a month)
- 1
- 000
- 000 tokens per month
- everything in Basic
- API access
- no daily limit
- priority GPU queue
- personal AI identity and agent API
- 5
- 000
- 000 shared tokens per month
- everything in Pro
- unlimited team seats
- daily
- weekly and monthly caps per member
- webhooks
- 99.9% uptime SLA
- reserved rate limits
- single invoice and volume-based token pricing
- 200K for USD 5
- 1M for USD 15
- 5M for USD 60
- 20M for USD 150
- 70M for USD 500
- 300M for USD 2
- 000
- Annual billing available on subscriptions at 20% off
Data, GDPR & hosting
A consolidated view of how Rewind.ai handles your data.
GDPR overview
GDPR coverage is partial and deliberately hedged. Asked “Is Rewind.ai GDPR compliant?”, the security page answers that the platform operates on GDPR principles — data minimisation, purpose limitation and the right to deletion — without ever stating compliance outright. A Data Processing Agreement is available on request. The privacy policy lists the expected rights: access, rectification, erasure, objection, portability and withdrawal of consent; its meta description also references GDPR and CCPA, and account deletion can be triggered from the settings, taking effect immediately and permanently, chat history and API keys included. What is missing is substantial: no Article 27 EU representative, no named data protection officer, no postal address, no declared hosting country or region — GPUs are only described as hosted at Vultr — and no published list of subprocessors.
Who owns the data?
The terms keep the service itself — website, branding and software — owned by Rewind.ai or its licensors and protected by intellectual property law. What users produce runs the other way: the FAQ and the about page state that generated outputs belong to the user, for personal or commercial work, and that every self-hosted model carries an MIT or Apache 2.0 licence, so results are reusable. Account details, prompts, uploaded files and outputs are collected by Rewind.ai to operate the service, without any ownership claim over them. Users remain responsible for the lawfulness of what they do with the outputs they generate.
Reuse rights
The privacy policy, effective 25 March 2026, lists four categories of collected data: account details (email address, password hash, profile settings), usage data, user content (prompts, messages, files and outputs) and payment data handled by Stripe. The stated purposes are limited to providing the service, maintaining and improving performance, preventing fraud and abuse, and communicating with users. Rewind.ai states that it does not sell personal data and shares it only with service providers — payment and infrastructure — or where legally required. User content is not used to train AI models unless the user explicitly opts in; any future opt-in training is promised to be clearly disclosed and off by default. Cookies are limited to strictly necessary, preference and analytics categories, and card details are never stored on Rewind.ai servers, Stripe being certified PCI Level 1. Users may reuse their own outputs freely, including commercially, without asking permission.
Data retention & training
Hosting summary
No hosting country or region is declared anywhere on the site. Inference runs on NVIDIA A100 GPUs described as hosted at Vultr, with no location given; the front end resolves behind Cloudflare (104.26.3.230, AS13335, anycast), which is a network observation rather than a statement by the publisher. The security page describes hardened VPS instances and cloud GPU machines, SSH access by key only with no password login, automatic security patching, restrictive firewall rules and encrypted database backups. Traffic is encrypted in transit over HTTPS with TLS 1.2 or above and HSTS enabled, including towards the inference servers, and requests processed by self-hosted models are said never to leave the operator’s own servers. Payments go through Stripe, certified PCI Level 1, so no card data sits on the publisher’s infrastructure. What is missing matters for compliance work: no jurisdiction, no data residency option, no subprocessor list and no third-party certification such as SOC 2 or ISO 27001. A Data Processing Agreement is available on request.
Things to keep in mind
Risks and trade-offs to weigh before adopting Rewind.ai.
- The operator is unidentified — no legal entity, no address, no named officer, and a single support address obfuscated in the page source — which makes recourse difficult if something goes wrong
- The domain was registered on 1 May 2026 with NameCheap: the platform is less than four months old at review, with no track record to lean on
- The name collides with the original Rewind AI: never assume the two products share anything, least of all their privacy commitments
- No explicit GDPR commitment, no EU representative and no declared data location — think twice before pushing personal or client data through the platform
- No subprocessor list is published even though Stripe and Vultr are named elsewhere on the site, and retention is left unquantified as “only as long as necessary”
- The site’s own figures disagree between pages, 400+ tools against 200+, a reminder that on-site counters such as 2K creators or 25.0M tokens cannot be verified
- AI outputs can be wrong and the terms place verification on the user; convenience at this scale also makes it easy to stop checking, to lean on the machine for thinking, and to let one free account quietly become the default for work it was never audited for
Setup & Integrations
Technical difficulty
Very low for web use: nothing to install, no account needed to start, and no configuration for any tool in the catalogue. A free account is created from an email address alone, with no initial password. The API path is barely harder for a developer: create an sk-rewind- key, point an existing OpenAI-compatible client at the Rewind.ai base URL and swap the key, the endpoints being compatible with no other announced code change, with cURL, Python and Node.js examples provided. Integrations are the only real effort: Slack, WordPress and Zapier go through webhooks or a shortcode, with no packaged connector.
Deployment
Integrations
Supported languages
Behind Rewind.ai
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Rewind.ai.
Frequently asked questions
Do I need an account to use Rewind.ai?
Is it really free?
Which models can I use?
Can I use the results commercially?
Is my data used to train AI models?
Is there an API?
Is there a mobile app or a browser extension?
Is this the original Rewind AI, with the Mac app and the Pendant?
What is the cheapest paid option?
Can I delete my account, and is a DPA available?
Should you pick Rewind.ai?
Rewind.ai has chosen breadth over depth: rather than excelling at one function, it covers most families of AI at close to no cost. The bet holds because the operator owns its GPUs and runs open-weight models on them, which is what makes a permanent free tier — 5,000 tokens a day, no account, no card — sustainable, while external models are passed through at provider cost.
What sets the platform apart is its technical candour. Few tools publish their stack, their GPU host, the licences of the models they serve and the token price of a request before you send it. That openness makes the free tier credible and the API easy to adopt: change the base URL and the key, and an existing OpenAI client works.
The symmetry is uncomfortable, because the legal side is exactly as opaque as the technical side is transparent: no legal entity, no postal address, no named officer, no hosting country, no subprocessor list, no certification, no explicit GDPR compliance statement and no quantified retention period. The domain was re-registered on 1 May 2026, so the platform is barely four months old at review, and its name carries the memory of a well-known product — the original Rewind AI Mac app and Pendant, later Limitless, bought by Meta and shut down in December 2025 — with which the site states it has no affiliation. The clarification is honest; the confusion will still linger.
The practical verdict follows from that split. Rewind.ai is a sound place to experiment, to work personally and for teams that do not push sensitive or regulated data through it. For procurement-grade requirements — a named counterparty, documented data residency, an EU representative — it is not a candidate yet.
- Choosing a selection results in a full page refresh.
- Opens in a new window.