Videos now auto-transcribe narration
A contractor uploaded a walkthrough video the other day, expected the AI to use his stated measurements, and got back a proposal that guessed every dimension from the frames. His narration — the most accurate dimension source you have — was silently dropped because the room photo analyzer doesn't listen to video audio.
That's now fixed for every video you upload.
What changed
Every video upload now extracts its narration in the background. When you film a walkthrough and call out "the back wall is five-eight, niche is mid-height on the right", BidCraft pulls the audio off the file, transcribes it, and saves it as a voice note attached to that room. Same as if you'd recorded the voice note directly — and voice notes are the highest-trust source the scope drafter has.
You'll see a small ✓ Narration pill appear on the video chip a minute or two after upload. The transcript lands in your Notes for that room.
Tap a video to play it
Video chips inside a room are clickable now. Tap any one and an inline preview panel opens with a playable video right inside the room view. You can verify the upload worked, watch what you actually captured, and see the transcription status — all without bouncing around.
When transcription doesn't go right
The transcription pipeline isn't magic. Sometimes a video really is silent (you didn't narrate), sometimes the audio track is too noisy or compressed, and sometimes the run just fails. Each case has a clear signal on the video chip:
- No audio — the AI couldn't find any speech. Try again with the Try again button in the preview, or add a voice or text note directly.
- Failed — the transcription run hit an error. Tap Retry transcription in the preview panel.
- No pill yet — either the transcription is still running, or the video predates this feature. Open the preview and use Extract narration to start it.
The retry is safe to tap as many times as you want — BidCraft won't create duplicate notes.
"Why can't I draft a scope yet?"
Tied to this: when you've uploaded inputs but the scope drafter can't work with them yet (a video that's still being transcribed, photos that haven't been analyzed, no contractor notes), the project page now tells you why — and what specifically to do next — instead of leaving Step 2 silent. Add a note, analyze your photos, or wait for narration to land. Step 2 lights up automatically once the gate's cleared.
Why this matters
Two-thirds of the proposal accuracy problem is having enough context for the AI. The fastest source of that context is you talking while you film. This change makes sure none of that gets thrown away.