AI research workspace

ResearchStudio

ONLINE20

Move research from
question to impact.

Find evidence, shape novel ideas, check claims, and turn finished work into publication-ready stories.

View on GitHub ↗

Explore alphaXiv ⌄

ResearchStudio-IdeaView on alphaXiv ↗ ResearchStudio-ReelView on alphaXiv ↗

Project Site ↗

Try it LIVE

Idea Spark Spark a novel idea 👆

Paper Searchsoon Scoop Checksoon

Experiment Code Analysis

Coming soon Autonomous experiment & code agents in development

Poster Generate Paper → poster 👆

Videosoon Blogsoon Reelsoon

Pick a skill, then add context below

START A RESEARCH RUN

Add your research context.

Pick a skill in the panel above, then fill in the details.

Research Problem Describe your research domain and the problem you want to explore. (Optional: reference arxiv links, compute budget, or other constraints.)

Claimed Novelty Your specific claimed contribution — a new method, theorem, dataset, framing, or insight — that sets this work apart from prior art. Paper PDF Upload the paper you want to transform. PDF only · max 50 MiB · 100 pages.

Paper source

arXiv PDF Title

arXiv id or link. We download the PDF. PDF · max 50 MiB · 100 pages. We search arXiv for the closest match.

Orientation

Landscape Portraitsoon

Layout

3-column 4-column Full Half

Method focusMethod figure + steps get their own columns

Block style

random solid framed simple left-bar elevated neo-brutal tag underline tinted double-rule legend-frame

Title banner

random Balanced Mirrored Logo strip Title-left Classic + QR

Color

random mono blue teal green burgundy purple rust slate plum

Advanced

Scan / QR

random single dual aside hero contact directory banner twin chips

Math

random KaTeX MathJax

Font

random Calibri Aptos Cambria Arial Times Verdana Georgia Trebuchet

Special requirements

Open ↗

Style preview only. Layout, block style, header, and color update live on a real poster skeleton; the text is blurred sample content and the conference tag is a placeholder. Your paper's real content fills it in when you generate.

Prefer to run it yourself? Copy the exact prompt

Exact paper2poster prompt for this configuration

These are the exact directives this studio would send. The first line points at the paper you entered (arXiv link, or your local PDF path); adjust it if needed, then paste into Claude Code (the paper2poster skill). Handy when the shared demo is busy: run it on your own machine with no queue.

Paper source

arXiv PDF Title

arXiv id or link. We download the PDF. PDF · max 50 MiB · 100 pages. We search arXiv for the closest match.

Target duration

Auto 3 min 5 min 8 min 10 min Custom

minutes Auto analyzes the paper's content, complexity, and narrative structure to choose the most appropriate duration. Selecting a specific duration makes the script and narration target that length.

Narration voice

Andrew (balanced) Aria (clear) Christopher (authoritative) Guy (energetic) Jenny (friendly)

These are real voices reported by the installed Edge TTS backend.

Subtitles On

Boxed Plain Bottom bar Toggleable (CC)

Boxed and Plain overlay the slide. Bottom bar scales the complete slide above a solid black caption band, so captions never cover PPT content. Toggleable embeds a soft caption track that compatible players can switch on or off.

Pointer style

Cursor Laser None

Selects the moving pointer that follows narration-aligned evidence.

Spotlight style

Spotlight Focus box None

Selects the region treatment around the evidence. Pointer and spotlight choices can be combined freely.

Resolution

720p 1080p 1440p 4K

Special requirements

PPT Options (from ppt-master) Deck design + editable PPTX export

CanvasPPT 16:9 · 1280 × 720locked for video

These selections are submitted as your confirmation for PPT-Master's Strategist gate. Auto means you pre-approve its paper-specific recommendation.

How PPT and subtitles work together Each slide is matched to one narration section, and word-level timestamps place each caption at the correct moment. PPT options change only the slide design; Subtitle mode controls how captions appear in the video.

Slide count One number or a range, up to 60. Narrative mode Controls how the presentation tells the story—for example, leading with the conclusion, explaining step by step, or following the paper's research journey. Visual style controls how it looks. Visual style

Target audience Leave blank = infer from the paper. Blank does not mean “no audience.”

Color direction Typography direction Leave blank = derive a PPT-safe font system from the style and audience.

Advanced PPT controls Images, formulas, and editable PPTX behavior

Image sources

Auto AI-generated Web-sourced Paper / provided Placeholder No images

Auto means PPT-Master chooses one source or an intentional mix for this paper; it does not always mean “mixed.” Auto and No images are exclusive.

Formula policy

PPTX editability Downloadable PowerPoint only. These settings do not affect narration or subtitles.

Charts & tables in PPTX The default keeps the closest visual match. Native objects let you edit chart/table data in PowerPoint, but may look slightly different. Text editing in PPTX Editable paragraphs are easier to revise. Exact line positions preserve the designed line breaks but make text editing less convenient.

preview

What this configuration controls

Paper source

arXiv PDF Title

arXiv id or link. We download and validate the PDF. PDF · max 50 MiB · 100 pages. We search arXiv for the closest match.

Target audience

Broad technical readers Domain researchers Industry practitioners

Controls terminology, explanation depth, and how much practical context the articles provide.

Editorial focus

Balanced Method Results Implications

Controls section order and relative depth without changing the paper's evidence or claims.

Figure use

Auto · 3–7 Focused · 3–4 Balanced · 4–5 Visual · 6–7

Only useful paper-native figures are included. The generator will not invent or duplicate images to reach a target.

Result table

Auto Prefer a table No table

A table is used only when exact results from the paper support it. Prefer never means fabricate.

Additional requirements

Output package

What you’ll receive

One evidence-aligned bilingual editorial package, written natively for each readership rather than translated sentence by sentence.

English research blogEditable blog_en.docx

中

Chinese WeChat articleEditable blog_zh.docx

The same claims, numbers, and selected paper figures in both versions
Localized titles, section copy, figure captions, and source links
An optional compact result table when the paper supports one
Embedded, cropped paper figures ready for an editing workflow
Strict bilingual, media, font, and rendered-layout quality checks

No live preview These controls change article content and structure. The actual Chinese and English articles are shown after generation.

Paper source

arXiv PDF Title

arXiv id or link. We download and validate the PDF. PDF · max 50 MiB · 100 pages. We search arXiv for the closest match.

Poster OptionsPaper2Poster

Orientation Layout

Method focusGive the method figure and steps dedicated columns.

Block style Title banner Color

Advanced poster controls

Scan / QR Math Font

Special requirements

Video OptionsPaper2Video + PPT-Master

Target duration Custom duration (minutes) Narration voice Resolution Pointer style Spotlight style

SubtitlesInclude subtitles in the delivered video. Subtitle style Special requirements

PPT Options (from ppt-master)Deck design

Slide count Narrative mode Visual style

Target audience

Color direction Typography direction

Advanced PPT controlsImages, formulas, editability

Image sources

Auto AI-generated Web-sourced Paper / provided Placeholder No images

Formula policy

PPTX editability Downloadable PowerPoint only.

Charts & tables in PPTX Text editing in PPTX

Blog OptionsPaper2Blog

Target audience Editorial focus Figure use Result table

Additional requirements

Assembly pipeline

Shared extractionPaper text, figures, tables, formulas, and metadata.
Full upstream artifactsPoster, bilingual blog, slides, narration, and video.
Section alignmentPoster regions are matched with the corresponding video and article sections.
Interactive ReelThe aligned assets are assembled into one self-contained viewer and download package.

Production viewer Paper2Reel uses its fixed poster-first interaction model. It does not expose a separate visual skin selector.

Interactive output

One paper, one navigable package

The result opens on the research poster. Double-clicking a poster section reveals the aligned video, subtitles, slides, and bilingual article content.

Poster-first view Research title

Double-click a section

→

Aligned section modal

▶Video · CC

EN / 中文Article + slides

AllZIP streamed on demand

PosterZIP streamed on demand

VideoZIP streamed on demand

BlogZIP streamed on demand

Downloads are packaged only when clicked and streamed directly to your browser; no ZIP archive is left in the Reel project.

Output bundle structure

reel.html
content_alignment.json
manifest.json
assets/
  poster/
  media/
  slides/
  blog/
  meta/
    reel_downloads.json

Viewer controls Top menu, audio, fullscreen, help, optional captions, English/中文 article switching, slide thumbnails, and a draggable video/article split.

Your Email We’ll send the finished report here. We don’t save email addresses.

Runs securely in the background. You can close this tab.

Research run queued

Access Token
Result URL

Keep this token safe — it’s the private key to your report.

ResearchStudio

Open-source AI workspace for the full research journey.

Project Site ↗ GitHub ↗ Idea Paper ↗ Reel Paper ↗

For research purposes only. Please do not abuse, scan, or attack this service.