The rendered flow matches the prompt’s full stage sequence and end branch: Document → Text Extraction → Chunking → Embeddings → split into Vector DB and Metadata Store.
What was measured
Text-to-animation accuracy
Does the output match all stages in the input description?
decisive for this rankingtransformation
This is the core requirement: the animation must match the described stages and sequence. (3 of 3 judges)
What was given, what came back
Test input: RAG Ingestion Pipeline · text
Input — what we sent
The exact prompt
Create an animated flowchart titled "RAG Ingestion Pipeline". A Document is uploaded and its text is extracted. The extracted text is split into smaller chunks. Each chunk is converted into embeddings. The generated embeddings are stored in a Vector Database, and the metadata is stored in a Metadata Store.
A text prompt asking a tool to create an animated flowchart of a RAG ingestion pipeline: document upload, text extraction, chunking, embeddings, and parallel storage into a vector database and a metadata store. It stresses sequential pipeline animation plus a branch into two output paths.
Why this input is hard
- · Sequential process visualization
- · Parallel branching into two outputs
- · Label accuracy for pipeline stages
- · Animated flowchart generation from plain text
Output — unretouched
Also checked on this input — same tool, 2 other criteria
Visual clarity⚠ StruggledFive of the six nodes render with empty placeholder glyphs instead of real icons; only Chunking shows the correct scissors icon.Visual clarity⚠ StruggledThe Metadata Store branch is visibly de-emphasized relative to the Vector DB branch, reading as a dim afterthought rather than an equally legible output path.
Provenance
- Observation
- 9b6b45ac-4e17-4a9e-9e40-c0d01ef8c364
- Evidence run
- e7fbb451-ba1e-4de0-ad94-41cde2700aef
- Study
- Generate Diagram Animations from Text Descriptions
- Research task
- 86b8vp172
- Tested at
- not recorded
- Source
- first-party
- Evidence state
- verified
- Proof shown
- input + output shown
- Cost / latency
- not captured
- Repeat run
- not captured
- Tester
- not captured
The last three rows are honest blanks, not placeholders — our capture has no field for them yet.
Query this
get_evidence({
tool: "remotion-ai"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Same input, same check — 9 other tools
measured on Text-to-animation accuracy
Academa AI✓ WorkedThe tool preserved the full six-step RAG ingestion sequence and the split into two parallel outputs, with document upload, text extraction, chunking, embeddings, vector database storage, and metadata storage all represented in order.AnimG✗ FailedThe tool can fail to keep parallel storage targets visually distinct; in the tested RAG flowchart, Vector DB and Metadata Store end up stacked on top of each other instead of separating into two nodes.Claude AI◐ MixedReproduces the requested ingestion pipeline structure and parallel branching correctly, including the split to both Vector Database and Metadata Store, but does not render the requested in-canvas title 'RAG Ingestion Pipeline'.EasyMotion✓ WorkedThe rendered flowchart matches the five-step RAG ingestion sequence and splits the final stage into two parallel destinations: document/input, extraction, chunks, embeddings, then vector DB and metadata storage.Framia✗ FailedThe recap screen can add hallucinated extra boxes and misstate the final structure by drawing Metadata Store and Vector Database as a sequential arrow instead of parallel destinations.Kodisc✓ WorkedRenders the full RAG ingestion sequence correctly, including document upload, text extraction, chunking, embedding generation, and fan-out to both a vector database and a metadata store.Replit✓ WorkedOn the tested prompt, the generated animation matched the requested pipeline structure exactly, including the split into both Vector Database and Metadata Store, and the report records zero spelling errors across all node labels.Vismo Studio✓ WorkedThe tool rendered the full 5-step ingestion sequence in order: document upload, text extraction, chunking, embedding generation, and parallel storage into both a vector database and a metadata store.X-Pilot✓ WorkedRenders the full RAG ingestion sequence correctly, including the sequential steps and the branch into both the Vector Database and Metadata Store.
This evidence is published in
Real inputs and real outputs, no retouching · every cell queryable via API & MCP · aidemos.com