Two wallets can propose the same Bitcoin payment and show different total fees, even at the same fee rate. One reason is the transaction each wallet has assembled underneath the amount field.
For readers comparing Bitcoin transaction size, three measurements need separating: serialized bytes, weight units and virtual bytes. This page follows a deliberately specified native SegWit example. It is about reading a proposed onchain transaction, not choosing a live fee rate or reorganizing a wallet's funds.
Bytes, weight and vbytes answer different questions
Serialized size counts the transaction's bytes, including witness data when present. Base size counts the serialization with witness-related data stripped. Witness data includes information used to satisfy an input's spending conditions, such as a signature and public key in our example.
BIP141 defines the accounting:
Weight = 3 × base bytes + total bytes
Virtual size = weight ÷ 4, rounded up to a whole vbyte
An equivalent weight expression is four times the base size plus the additional witness-related bytes. Bitcoin Core's v31.1 weight implementation uses serialization with and without witness data to calculate it. A virtual byte, abbreviated vB, is an accounting unit; it is not another name for one byte in the complete serialized transaction.
The 31.0 decoder reference lists size, weight and vsize separately. A screen reporting 223 bytes and 141 vB can be internally consistent. Those fields are measuring different things.
Declare the transaction before calculating its size
An input spends a particular earlier output. More selected inputs mean more serialized data, but the exact contribution depends on how those outputs are spent. A wallet balance alone does not specify the input count.
For our fictional example, assume:
- Native pay-to-witness-public-key-hash, or P2WPKH, inputs with empty
scriptSigfields. - A 72-byte signature item, including its sighash-type byte, and a compressed 33-byte public key for each input.
- Two P2WPKH outputs: one recipient payment and one change output. Each output script is 22 bytes.
- Four-byte version and locktime fields; the small counts and lengths here each fit in one CompactSize byte.
- No other inputs, outputs or optional data. The selected inputs can be spent, and this is not a coinbase transaction.
These assumptions follow BIP141's native P2WPKH structure and the v31.1 transaction serialization. They are not universal constants for every signature, script or wallet. No signed transaction has been constructed or broadcast for this article.
One input and two outputs: 141 vB in this example
First count the base bytes:
| Component | Declared calculation | Base bytes |
|---|---|---|
| Version, input count, output count, locktime | 4 + 1 + 1 + 4 | 10 |
| One input | 36-byte outpoint + 1-byte empty-script length + 4-byte sequence | 41 |
| Two outputs | 2 × (8-byte value + 1-byte script length + 22-byte script) | 62 |
| Base serialization | 10 + 41 + 62 | 113 |
The input's witness occupies 108 bytes: one stack-item count byte, one signature-length byte, the assumed 72-byte signature, one public-key-length byte and the 33-byte public key. The complete SegWit serialization also has a two-byte marker and flag.
Total size is therefore 113 + 108 + 2 = 223 bytes. Weight is 3 × 113 + 223 = 562 WU, where WU means weight units. Dividing 562 by four gives 140.5; rounding up gives 141 vB.
Do not round each component separately and then assume its rounded contributions add to the same result. Here, rounding applies to the transaction's combined weight.
The same payment with two inputs costs more at the same rate
Now compare two alternative transactions. Both draw on 100,000 sats in total and send 50,000 sats to a recipient. Alternative A spends one 100,000-sat output; B spends two 50,000-sat outputs. Each creates the same two types of new outputs: recipient and change.
For B, base size becomes 10 + 2 × 41 + 62 = 154 bytes. Total size is 154 + 2 × 108 + 2 = 372 bytes. Weight is 834 WU, giving 209 vB after rounding up.
Use an invented rate of 5 sat/vB for both. This is not a current fee quote, a recommended setting or a promise of inclusion.
| Measurement | A: one input | B: two inputs |
|---|---|---|
| Input value total | 100,000 sats | 100,000 sats |
| Recipient payment | 50,000 sats | 50,000 sats |
| Virtual size | 141 vB | 209 vB |
| Assumed fee rate | 5 sat/vB | 5 sat/vB |
| Calculated total fee | 705 sats | 1,045 sats |
| Change output | 49,295 sats | 48,955 sats |
The difference is 340 sats because B uses 68 more vB at the assumed rate. The amount paid to the recipient did not change. These are alternatives, not two payments sent sequentially.
The fee also reconciles as total input value minus total output value. For A: 100,000 − 50,000 − 49,295 = 705 sats. There is no third "fee output." Our change-address explainer owns the separate question of where the remainder goes. The Developer Guide supplies the underlying input/output accounting, not a current fee threshold.
Why another transaction can have a different answer
Change the signature length, input type, output scripts or number of outputs and the size calculation can change. A transaction with no change output is different from either example. Larger counts can also need longer CompactSize encodings. "One Bitcoin transfer" is not a fixed byte length.
There is a policy distinction too. Core's v31.1 policy implementation can adjust virtual size upward when signature-operation cost dominates. Our ordinary P2WPKH examples use the weight-based result; they do not model a sigop-heavy script or establish every node's admission policy.
More inputs can raise the present transaction's fee, but that observation is not an instruction to consolidate funds. Combining outputs has separate fee, privacy and future-spending consequences. This explainer changes no wallet settings and prescribes no input-selection strategy.
Check fee units before comparing estimates
Total fee is an amount: sats. Fee rate is an amount per virtual byte: sat/vB. At the same rate, a larger transaction pays a higher total fee. A larger absolute payment does not automatically make its serialized value field longer.
Interfaces need reading on their own terms. The 31.0 estimatesmartfee documentation reports feerate in BTC/kvB, not sat/vB, and describes an approximate estimate that can be unavailable. One kvB means 1,000 vB. A bare number without its unit is not a usable comparison.
An estimate does not reserve space in a block. RBF and CPFP concern changing incentives for pending transactions; they are not size measurements or guaranteed cancellation tools. Confirmations and exchange credit remain separate from a wallet's estimated fee.
A useful transaction-size record
Keep the input count and spending types, output count and scripts, complete serialized size, weight and virtual size together. Record the fee unit and total fee, then reconcile the input and output values. If the transaction is still unsigned, distinguish an estimate from the final signed size.
Bitcoin transaction size helps explain the fee without turning the payment amount into a percentage charge. The useful comparison is the actual structure and its fee rate, with assumptions visible. Same payment, different packing list.
Sources
Primary specifications, source and exported documentation checked October 27, 2026, Asia/Tbilisi. Core code is pinned to v31.1; RPC references are explicitly 31.0. The Developer Guide contains historical policy and fee passages, which are not used as current rules. The examples and illustration are fictional. No node, wallet, RPC, fee estimator, signature or transfer was tested.
- BIP141: serialization, P2WPKH, weight and virtual size.
- Bitcoin Core v31.1: transaction serialization.
- Bitcoin Developer Guide: transaction inputs, outputs and fee accounting.
- Bitcoin Core 31.0: decoder size, weight and vsize fields.
- Bitcoin Core 31.0: fee-estimate units and limitations.
- Bitcoin Core v31.1: sigop-adjusted virtual-size policy.
- Bitcoin Core v31.1: transaction weight calculation.
See something wrong? Brokzi logs material changes. Prepare a correction note.




