User Manual

IDEAINNOV (ideainnov.com) is a research SaaS offering 4 automatic-generation programs for researchers. The current version is V140. After signing in to the dashboard, pick a program from the 4 cards.

The chatbot assistant is named “unifi” (통일 in Korean). The name reflects the goal of integrated research across literature, philosophy, psychology, economics, engineering, chemistry, and more. You can ask it anything by voice or text via the widget in the bottom-right corner, or on /help.

The 4 programs

• 🧬 Paper Writing — V140 Paper Pipeline: turns a natural-language topic or a single source paper (PDF/DOCX/HWPX) into an SCI-Q1-grade academic paper draft. Written in SCI format (Introduction, Related Work, Hypotheses, Method, Experiments, Results, Discussion, Conclusion), with editable numbered OMML equations, figures and tables, and automatic citations (measured counts per public sample are on the /paper page). Accuracy is raised by our own coupled-equation technology. Several reviewers cross-examine the draft against SCI Q1 criteria; the download link stays valid for 24 hours on the free trial, and paid plans keep the output in your archive for 30 days.
• 💼 Business Proposal Writing — buz_engin: auto-generates PR/marketing proposals, campaign content, and a PPTX deck from an RFP or a marketing topic. Extracts 10 core keywords to design target/channel/message strategy, derives 3 Win Themes and KPIs, and produces a 40–80 slide deck in the Impact-8 structure.
• 📑 HWPX Conversion — HWPX ↔ DOCX ↔ PDF Converter: converts between Hangul (HWPX), MS Word (DOCX), and PDF while preserving style, equations, and tables. Equations become editable OMML objects; visual fidelity (SSIM) is measured against the original.
• 📤 Publishing System — SCI Journal Submission Matching: recommends suitable SCI journals from your title/abstract/keywords and prepares you up to the portal step. See §4.

Pricing / credits summary

• 12 free credits for each of the 3 programs on sign-up (36 total).
• Credits per generation: paper regular 1, hard mode 2 (scales with iteration count), BIZ proposal base 1 (+1 per extra option), HWPX conversion 2.
• Upload/conversion file limits: Free/Basic 20MB·30p · Pro 40MB·100p · Enterprise 100MB·200p.
• Refunds: unused credits/subscriptions are fully refundable within 7 days of payment (14 days for overseas Paddle payments). See /billing for the plan comparison and /terms for the policy.

Several accounts on the same network (IP)

• Several accounts may sign up and sign in from the same public IP (a shared home, office or campus router) — no account is blocked. The free trial, however, is one pool shared per IP: accounts on the same IP together share 1 person's worth (36 credits in total). When that pool runs out, only free runs stop; the accounts keep working and paid plans are available as usual.
• Accounts with a university or research-institute email (*.ac.kr, *.re.kr, *.edu, *.edu.xx (xx = two-letter country code)) each get their own free pool on the same IP up to the 15th account; from the next account on, the shared IP pool rule applies.
• Paid accounts (an active subscription or a paid credit balance) are exempt from every limit in this IP policy.
• Signing in from a network (IP) different from the one you first used asks you to re-verify once with an email code. Verified extra IPs are remembered up to 10, most recently used first; beyond that, the least recently used IP is dropped.

Desktop app (beta)

IDEAINNOV Desktop (Windows 10/11, 64-bit) stores your query history locally on your PC rather than the server, auto-organizes and verifies (sha256) your personal output vault, and keeps your login session persistent. It's currently in beta — email info@ideainnov.com to request access. See /desktop for details.

Security / IP principle: internal algorithm parameters, equations, and source code are not disclosed. We explain WHAT a feature does, but not HOW it is implemented (internal numbers/code/algorithm names).

Basic procedure

1. Enter a research topic as a natural-language sentence, or upload a source paper (PDF/DOCX/HWPX).
2. The system analyzes the query, auto-classifies the domain, and derives a matching coupled PDE system.
3. Several reviewers cross-examine the draft against SCI Q1 criteria and improve it, and it passes one more final check right before delivery (see §3).
4. Download the finished DOCX (editable OMML equations, figures and tables). The download link stays valid for 24 hours on the free trial; paid plans keep it in your personal archive for 30 days.

Tips for writing a good query

A good query makes a good result. Following these principles noticeably improves quality.

• State your domain clearly. Especially for non-physics/engineering fields (economics, business, social science, operations management, etc.), start the sentence with “This is a [field] problem” for a more accurate result.
• Quantify conditions with numbers. E.g. “flow variation ±20%”, “temperature range 20–80°C”, “sample size n=50”. Besides IC/BC, any quantitative operating condition you specify will appear in the body.
• Keep the core idea to one sentence. A short, clear query gives the most reliable result.
• State a comparison baseline if you have one. “Compared to the existing method”, “compared to prior work X” — this helps the Related Work section.

Language (Korean / English) choice

Both Korean (KO) and English (EN) are supported. Korean generation may take longer than English (we have no measured per-language comparison yet). We're being upfront that Korean jobs can take longer than the on-screen estimate suggests. If you're in a hurry, generating in English first and translating separately is one option.

File upload

Supported formats are PDF, DOCX, and HWPX; upload size/page limits vary by plan tier (see §1). If you upload a source paper, the system works toward improving/extending that paper's governing equations.

Iterative interpretation (연속 심화 해석)

Re-solves the same problem over multiple rounds, each round starting from the previous solution to refine the equations further. Accuracy (R²) improves with each round but gradually converges (e.g. single-physics baseline R²≈0.982 → iterated coupled R²≈0.995). Up to 3 rounds are allowed, and credits scale with the round count (e.g. hard mode × 3 rounds = 6 credits).

Hard-problem mode

A deep-research option for genuinely difficult engineering/math/science problems. When enabled, it derives deeper governing equations, solves them as a coupled system, and applies several rounds of cross-review validation. It costs more credits than regular mode (2 vs 1), so we recommend using it only for genuinely hard problems.

Honest expectations on processing time

Paper generation takes a measured median 94 min (production server, n=3, 85–151+ min). Measured on 2026-08-31 by submitting the same English query 3 times to the production server; the longest run was still processing at 151 min, so it actually took longer. With fewer than 30 runs we report the median and range, not an average. It goes through domain classification → equation derivation → multi-stage validation → document assembly, so it doesn't complete instantly. Hard mode and more iteration rounds add stages and take longer, and Korean output may also take longer. You'll be notified by email when it's done even if you leave the page, and the result is saved to your personal archive.

The “coupled PDE prediction” the system produces is fundamentally an internal simulation result. To check how well it matches reality, it has to be compared against real measured data obtained outside the system — a process called External Validation. A label appears in the paper to disclose whether and how far this was done. Below we explain each label plainly.

Validation grade printed in the paper

The paper states which of three grades of evidence backs its results. Same-condition measured means the result was compared directly with public measured data taken under the same conditions the paper studies. Similar-condition substitute means no measured data exist for the same conditions, so public measured data from physically similar conditions were used instead; the paper qualifies this as a “substitute condition” and says how it differs from the original. If only some of the checks pass, the paper reports a “partial success”. For example, if the decay with distance and the overall magnitude match but the error is not clearly smaller than the baseline equation, the broad form of the equation holds while the effect of the newly added term is not demonstrated. Passing most checks is reported as “supported” and passing few as “refuted”, exactly as found. Synthetic means no measured data could be used and only results computed inside the system exist; accuracy figures then carry a (synthetic) mark and the Limitations section says so. Please check this grade before you submit.

If you have questions, contact info@ideainnov.com anytime.

Just enter your paper's title, abstract, and keywords to get the top 10 candidate SCI journals for it, complete with direct links to each journal's submission portal — helping you go from candidate search all the way to right before the portal step.

How to use it

1. Go to /submit (also reachable from the dashboard once signed in).
2. Enter your paper's title, abstract, and keywords.
3. The system recommends the top 10 candidate journals.
4. For each candidate, see publisher, quartile (Q1–Q4), SJR, APC, median review time, acceptance rate, and WoS Master (SCI)/Scopus/OA status.
5. Narrow candidates with filters: WoS Master (SCI) required, Q1 only, APC budget cap, etc.
6. Once you pick a journal, follow the provided link straight to its submission portal (ScholarOne, Editorial Manager, etc.).

Important: you submit the paper yourself

The system only automates candidate matching, information enrichment, and portal preparation. To comply with each journal's Terms of Service, you must click the final Submit button yourself on each journal's portal. The system never submits a paper on your behalf.

Recommended workflow

• We recommend drafting a paper first, then feeding the finished abstract/keywords straight into /submit to search for journals.
• If you're targeting a Q1 journal, turn on the “Q1 only” filter first, set an APC budget cap, then compare the remaining candidates — the most efficient order.
• WoS Master status directly affects whether your institution/department's SCI-indexing requirement is met, so check that filter if you have such a requirement.

The system applies one of three different approaches depending on the nature of the problem domain. Below is a user-level explanation of each; the actual before/after status across 30 research hard problems is shown as a table on the Hard-Problem Solving Showcase page (/showcase).

Physics / engineering — coupled-PDE approach

Problems involving continuously varying physical quantities — turbulence, superconductivity, plasma, biological signals — are expressed as a coupled system of partial differential equations (PDEs) linking multiple governing equations, solved together. The shape of the equations is progressively refined so it describes the real phenomenon more accurately.

Discrete mathematics — probability/visualization/combinatorial-optimization approach

Discrete problems that don't naturally fit a continuous equation — prime distributions, the repeating behavior of integer sequences, the combinatorial structure of graphs — use a separate, specialized approach instead: quantifying trends with probability distributions, revealing structural patterns through visualization, and searching for candidate solutions with combinatorial optimization. This is why these problems don't simply reuse the physics-family coupled-equation approach.

Social science — methodology-framing approach

Fields that rely on observational, survey, or policy data rather than controlled experiments — economics, business, social science — are approached not by deriving equations the way the physics family does, but by structuring the problem within that field's established research-methodology framework (hypothesis → operationalizing variables → analysis design → stating limitations). Stating your field at the start of the query (e.g. "This is an economics problem") helps produce a better-fit result in this case (see §2, Tips for writing a good query).

All three approaches produce a draft that a human must still review; they do not unconditionally guarantee accuracy or reproducibility. Internal algorithm parameters and equations are not disclosed.

Answers to questions users commonly ask.

The draft is cross-examined by several reviewers and passes one more final check right before it is delivered. Here is what you can verify in the output yourself.

Reference existence check

Every reference printed in the paper is looked up against external scholarly databases to confirm the source really exists. Anything that cannot be confirmed is replaced with a confirmed source on the same topic, or removed together with its in-text citation marker, after which the numbering is tidied up. The check can only confirm that “a source with this title exists”, so please make the final judgment on whether the citation fits your context.

Equations, figures and tables

Equations arrive as equation objects you can edit directly in Word, not as images. Standalone display equations are numbered, right-aligned, and cited by that number in the body. Every equation, figure and table is checked for an accompanying explanation in the body; anything left unexplained is fixed before delivery.

Appendix reproduction code

The reproduction code in the appendix is actually run to confirm it finishes without error, and its printed numbers are compared against the numbers written in the body.

Generative-AI use disclosure

A paragraph stating that generative AI was used is included in the Methods or Acknowledgements / Author Contributions section, in the form publisher AI-disclosure guidance asks for (for example the policies of Oxford University Press journals). It is good practice to state the same in your cover letter when you submit.

Wording diversity across papers

Sentence structures are varied so that papers produced by the same system do not read alike.

How these checks work internally, and what criteria they use, is not something we walk through in detail. What we can share is the behavior you can observe in the output.

• It never submits for you. SCI Submit automates candidate matching, enrichment and portal preparation only; you must press the final Submit button yourself on each journal portal, and complying with each publisher’s terms is your responsibility.
• External comparison against measured data is not always performed. If no public measured dataset can be obtained, the paper states that plainly and is written from internal analysis only. Whether a holdout (setting aside part of your own data for validation) or external validation was performed, and its scope, is stated in the paper’s labels — see §3 for how to read them.
• It is an AI-generated draft and needs human review. Numbers, citations and interpretation must be checked and revised by the researcher; final rights to and responsibility for the output belong to you.
• Places that need your own judgement are marked in the text with a red underline. Points that cannot be corrected automatically — where a number must change, a claim must be withdrawn, or only the researcher can decide — are marked on the exact sentence or number in red with a single underline. A one-line legend appears at the top of the document and an item-by-item review list at the end, saying what you need to check for each. The number of marks per paper is capped; anything beyond the cap is listed at the end without an underline in the body. A passage without a mark is not thereby certified correct. Anything already corrected automatically is not marked.
• Accuracy and reproducibility are not unconditionally guaranteed. We do not claim things like “it always improves by X%”.
• Neither acceptance nor publication is guaranteed.
• Internal implementation is not disclosed. The models used, internal judging criteria and algorithm details are not shared.

Recent updates

• SCI Submit added, free of charge. Get the top 10 candidate journals from your title, abstract and keywords (no credits; sign-in only).
• Document conversion in 10 directions. Converts across HWP/HWPX, DOCX and PDF (/convert).
• News tab added. Announcements and events now live at /news.
• 11-language interface. Korean, English, Japanese, Chinese, Spanish, French, German, Russian, Portuguese, Arabic and Polish — and the chatbot answers in the language you pick.
• “Run now” button. One click in the header takes you straight to the run screen on the dashboard.
• App install asks first. Installing the site to your home screen (PWA) does not proceed until you press Install on the confirmation card (on iPhone Safari, use “Add to Home Screen” from the share menu).
• Stronger paper quality checks. Reference existence verification, generative-AI use disclosure, numbered display equations cited in the body, an explanation for every equation/figure/table, appendix code execution and cross-check, and cross-paper wording diversity are now delivery conditions (§7).
• Several accounts per network (IP) allowed. Other accounts on the same IP are no longer blocked; only the free trial is shared as one pool per IP. University/research-institute accounts and paid accounts follow separate rules (§1).
• Validation grade in the paper. The paper now states whether its evidence is same-condition measured data, a similar-condition substitute, or synthetic only (§3).

Paper Challenge event (running now)

• Generate a draft with ideainnov.com Paper, finish the manuscript with your own judgment and edits, get journal recommendations from SCI Submit, and complete an actual submission — a total of 1,000 credits is awarded.
• Prizes (free credits): 1st 500 · 2nd 300 · 3rd 200.
• Entries close 2026-10-31; winners announced 2026-11-15 on the homepage.
• How to enter, in 4 steps:
  1. Use ideainnov.com Paper to generate a paper draft.
  2. Apply your own judgment and edits to produce your submission-ready DOCX manuscript.
  3. Use ideainnov.com SCI Submit to get journal recommendations and submit your paper.
  4. Email the materials below to info@ideainnov.com by 2026-10-31.
• What to submit: your judgment/edit notes, screenshots of the SCI Submit screen and its recommendations, the questions/options/files you input, and the final submitted output — email info@ideainnov.com.
• Judging criteria: completeness of the final paper, and the quality of your own interventions.
• In addition, the first 1,000 users to sign up receive 9 free credits.
• Full announcement: /news/paper-challenge-2026