FitFo
Turn fitness videos into workouts you actually do.
See a workout on TikTok or Reels? Share it straight to FitFo, the same way you'd send it to a friend. Our AI parses the video, pulls the exercises, sets, and reps, and builds a clean workout you can follow, edit, and track.
Built for people who scroll fitness content and want to actually train off it. Share a video, get a workout, track your progress - no ads, no social feed, just training.
~30s
To parse a video
100%
Private to you
0
Ads forever
How it works
From inspiration to execution, in three steps.
Share a fitness video
Hit the share button on any public TikTok or Reel and pick FitFo. No copy-pasting, no leaving the app you were scrolling.
AI does the work
Exercises, sets, reps, rest, notes - all extracted into a card you can edit, save, or start immediately.
Train it, log it, repeat
Follow the workout in-app, log every set, and come back to the same plan any day of the week.
Inside the app
Every screen is production. Here's what FitFo looks like in practice.
Import Workout
Paste a TikTok or Instagram Reel link and FitFo extracts the workout structure - exercises, sets, reps, and notes. You can also create a workout manually if you prefer.
Workout Detail
Every imported workout becomes a structured card - title, description, exercise count, muscle groups, and a direct link back to the source video. Start a session or save it for later.
Your Workout Hub
All your saved workouts in one place, filterable by muscle group - Chest, Back, Shoulders, Arms. Each card shows the source, exercise count, and lets you start a session or unsave in one tap.
Schedule the Week
Drop any saved workout onto a day. Get a clean calendar view of what's coming so you stop negotiating with yourself every morning. Tap a day to see what's planned.
Training Archive
Every rep, on the record. Completed sessions turn into a clean archive - see how many workouts and sets you've logged this month and schedule your best workouts again with a tap.
A RAG-grounded, single-creator fitness coach chatbot.
FitFo Coach answers training questions in one coach's actual voice, grounded in their TikTok corpus instead of generic model memory. Inline citations let users verify any cue, claim, or programming recommendation against the original source video.
The problem
A coach bot that cannot hallucinate the coach.
Architecture
TikTok profile -> Apify crawl -> Whisper transcript -> LLM chunker -> LLM tagger -> OpenAI embeddings -> pgvector HNSW search -> grounded LLM synthesis -> cited markdown answer
Discover
Apify crawls a creator's TikTok profile and upserts videos by platform ID for deduped ingestion.
Transcribe
TikWM, ffmpeg, and Whisper turn each source video into reusable transcript text with capped async concurrency.
Chunk
gpt-4.1-mini splits transcripts into self-contained training snippets and strips sponsor reads or pure CTAs.
Tag
A batched LLM pass assigns locked muscle, goal, exercise, and equipment tags that match SQL constraints.
Embed
text-embedding-3-small vectors land in pgvector with the model stored per row for future re-embedding.
Postgres-native retrieval
Supabase Postgres + pgvector keeps storage, tag filters, approval status, HNSW cosine search, and citations in one source of truth.
Filter before ranking
The match_content_chunks RPC narrows by creator, approval status, muscle groups, and goals before ranking top-K chunks by vector distance.
Grounded generation
Weak matches are dropped, empty retrieval skips the LLM, and every coaching claim must cite a retrieved chunk like [1].
What makes it production-ready
What you get
A training app built for content scrollers.
Import
Share a video, get a workout
Hit share on any TikTok or Reel and send it straight to FitFo. Our AI pulls the audio, reads the on-screen text, and turns it into a clean, structured session.
Organize
A library that knows what it is
Saved workouts, scheduled sessions, and logged history - each one tagged by muscle group and block so you can find them fast.
Train
Edit, follow, log - no friction
Tap any field to change reps, weights, or notes. Start a session, log every set, and the next one opens automatically.
Calendar
Schedule the week. Show up to it.
Drop any saved workout onto a day. Get a clean calendar view of what's coming so you stop negotiating with yourself every morning.
Logs
Every rep, on the record.
Completed sessions turn into a clean archive. See how many sets you've logged this month and schedule your best workouts again with a tap.
Built with
Shipped alongside a crew that trains.
nunoliftz
Co-builder
Jacob Oestreicher
Co-builder
Nirv
Co-builder
Technical Highlights
What powers FitFo under the hood.
Built AI video parsing pipeline that extracts exercises, sets, reps, and rest periods from TikTok and Instagram Reel content - handles spoken audio, on-screen text, and visual cues
Designed native iOS app with Swift - workout library, calendar scheduling, session logging, and muscle group tagging with zero-friction UX
Integrated iOS Share Extension so users can send videos directly from TikTok or Instagram to FitFo without leaving the app they were scrolling
Built workout scheduling system with calendar view - drop any saved workout onto a day and see your training week at a glance
Designed training archive with session history, set counting, and one-tap re-scheduling for completed workouts
Live on iOS
Train with the content you already love. FitFo is available now on the App Store. Share your first TikTok or Reel and be lifting off a real plan in under a minute.
Questions? nirv@fitfo.app