TikTok comments contain unfiltered brand sentiment: users asking 'where can I buy this', mentioning competitors, or complaining about features. This tutorial builds a comment mining pipeline using Scavio's video/comments endpoint to extract purchase intent, brand mentions, and product feedback at scale.
Prerequisites
- Python 3.8+
- requests library
- A Scavio API key from scavio.dev
- TikTok video IDs to analyze
Walkthrough
Step 1: Fetch comments from a video
Pull comments with pagination.
Python
import os, requests, re
from collections import Counter
API_KEY = os.environ['SCAVIO_API_KEY']
H = {'Authorization': f'Bearer {API_KEY}', 'Content-Type': 'application/json'}
def get_comments(video_id, pages=3):
comments = []
params = {'video_id': video_id, 'count': 20}
for _ in range(pages):
resp = requests.post('https://api.scavio.dev/api/v1/tiktok/video/comments',
headers=H, json=params).json()['data']
for c in resp.get('comments', []):
comments.append({'text': c.get('text', ''), 'likes': c.get('digg_count', 0)})
if not resp.get('has_more'): break
params['cursor'] = resp.get('cursor', 0)
return commentsStep 2: Detect brand mentions and intent
Scan for brand names and buying signals.
Python
BRANDS = ['scavio', 'serpapi', 'exa', 'tavily']
INTENT = ['where to buy', 'link', 'how much', 'price', 'worth it', 'recommend']
def analyze(text):
t = text.lower()
return {
'brands': [b for b in BRANDS if b in t],
'intent': [s for s in INTENT if s in t],
'negative': any(w in t for w in ['scam', 'fake', 'waste', 'terrible']),
}Step 3: Aggregate across videos
Analyze multiple videos and produce a summary.
Python
def report(video_ids):
for vid in video_ids:
comments = get_comments(vid)
mentions = Counter()
intent_count = 0
for c in comments:
a = analyze(c['text'])
for b in a['brands']: mentions[b] += 1
if a['intent']: intent_count += 1
print(f'Video {vid}: {len(comments)} comments')
print(f' Brands: {dict(mentions)}')
print(f' Intent signals: {intent_count}')
print(f'Cost: ${len(video_ids) * 3 * 0.005:.3f}')
report(['vid1', 'vid2'])Step 4: Export timeline data
Track brand signals over time.
Python
import json
from datetime import date
def export(video_ids, output='signals.jsonl'):
with open(output, 'a') as f:
for vid in video_ids:
comments = get_comments(vid, pages=2)
mentions = Counter()
for c in comments:
for b in analyze(c['text'])['brands']: mentions[b] += 1
f.write(json.dumps({'date': date.today().isoformat(), 'video': vid,
'mentions': dict(mentions), 'total': len(comments)}) + '\n')
print(f'Exported {len(video_ids)} videos')
export(['vid1'])Python Example
Python
import os, requests
from collections import Counter
API_KEY = os.environ['SCAVIO_API_KEY']
H = {'Authorization': f'Bearer {API_KEY}', 'Content-Type': 'application/json'}
BRANDS = ['scavio', 'serpapi', 'exa']
def analyze(video_id):
resp = requests.post('https://api.scavio.dev/api/v1/tiktok/video/comments',
headers=H, json={'video_id': video_id, 'count': 20}).json()
mentions = Counter()
for c in resp['data'].get('comments', []):
for b in BRANDS:
if b in c.get('text', '').lower(): mentions[b] += 1
print(f'{video_id}: {dict(mentions)}')
analyze('sample_id')JavaScript Example
JavaScript
const API_KEY = process.env.SCAVIO_API_KEY;
const H = { Authorization: `Bearer ${API_KEY}`, 'Content-Type': 'application/json' };
async function analyze(videoId) {
const resp = await fetch('https://api.scavio.dev/api/v1/tiktok/video/comments', {
method: 'POST', headers: H, body: JSON.stringify({ video_id: videoId, count: 20 })
}).then(r => r.json());
const mentions = {};
for (const c of resp.data.comments || []) {
for (const b of ['scavio', 'serpapi']) if (c.text?.toLowerCase().includes(b)) mentions[b] = (mentions[b]||0)+1;
}
console.log(`${videoId}:`, mentions);
}
analyze('sample_id').catch(console.error);Expected Output
JSON
Video vid1: 58 comments
Brands: {'scavio': 3, 'serpapi': 1}
Intent signals: 8
Video vid2: 42 comments
Brands: {'exa': 2}
Intent signals: 5
Cost: $0.030